CN111221896A - User behavior prediction method and device, electronic device, storage medium - Google Patents
User behavior prediction method and device, electronic device, storage medium Download PDFInfo
- Publication number
- CN111221896A CN111221896A CN201811428883.0A CN201811428883A CN111221896A CN 111221896 A CN111221896 A CN 111221896A CN 201811428883 A CN201811428883 A CN 201811428883A CN 111221896 A CN111221896 A CN 111221896A
- Authority
- CN
- China
- Prior art keywords
- long
- vector
- short
- time memory
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Theoretical Computer Science (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Neurology (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本公开提供了一种用户行为预测方法，涉及大数据技术领域。该方法包括：获取待预测用户在最近N个长周期内的行为数据，并转换为具有先后顺序的N个特征向量；将第一特征向量输入一长短时记忆网络模型，得到第一长短时记忆向量；依次地，根据第i‑1长短时记忆向量更新长短时记忆网络模型的初始记忆状态，并将第i特征向量输入长短时记忆网络模型，得到第i长短时记忆向量，其中i∈[2,N‑1]；根据第N‑1长短时记忆向量更新长短时记忆网络模型的初始记忆状态，并将第N特征向量输入长短时记忆网络模型，得到输出向量；根据输出向量确定待预测用户的行为预测结果。本公开可以挖掘用户行为的长期特征与变化趋势，提高用户行为预测的准确率。
The present disclosure provides a user behavior prediction method, which relates to the technical field of big data. The method includes: acquiring behavior data of the user to be predicted in the last N long periods, and converting them into N eigenvectors with sequential order; inputting the first eigenvectors into a long-short-term memory network model to obtain the first long-short-term memory vector; in turn, update the initial memory state of the long-short-term memory network model according to the i-1th long-short-term memory vector, and input the i-th feature vector into the long-short-term memory network model to obtain the i-th long-short-term memory vector, where i∈[ 2, N-1]; update the initial memory state of the long-short-term memory network model according to the N-1 long-short-term memory vector, and input the N-th feature vector into the long-short-term memory network model to obtain an output vector; determine the to-be-predicted according to the output vector User behavior prediction results. The present disclosure can mine long-term features and changing trends of user behaviors, and improve the accuracy of user behavior prediction.
Description
Technical Field
The present disclosure relates to the field of big data technologies, and in particular, to a user behavior prediction method, a user behavior prediction apparatus, an electronic device, and a computer-readable storage medium.
Background
With the development of big data technology, more and more enterprises are dedicated to predicting the behaviors and demands of users through big data analysis, so as to provide better services to attract more users.
Most of the existing user behavior prediction methods extract features from historical data of users, and behavior prediction results of the users are obtained by analyzing the features. Wherein, which features to extract depends on the experience of data analysts, so the validity of the features is difficult to guarantee; more historical data are recent data, and long-term characteristics of user behaviors and change trends of behavior habits cannot be reflected; the two defects result in low accuracy of user behavior prediction and influence on the effect of practical application.
It is to be noted that the information disclosed in the above background section is only for enhancement of understanding of the background of the present disclosure, and thus may include information that does not constitute prior art known to those of ordinary skill in the art.
Disclosure of Invention
The present disclosure provides a user behavior prediction method, a user behavior prediction apparatus, an electronic device, and a computer-readable storage medium, thereby overcoming, at least to some extent, the problem of low accuracy of the existing user behavior prediction method.
Additional features and advantages of the disclosure will be set forth in the detailed description which follows, or in part will be obvious from the description, or may be learned by practice of the disclosure.
According to an aspect of the present disclosure, there is provided a user behavior prediction method, including: acquiring behavior data of a user to be predicted in the latest N long periods, and converting the behavior data into N feature vectors with a sequence; inputting the first feature vector into a long-short time memory network model to obtain a first long-short time memory vector; sequentially updating the initial memory state of the long-short-time memory network model according to the i-1 th long-short-time memory vector, and inputting the i-th feature vector into the long-short-time memory network model to obtain the i-th long-short-time memory vector, wherein i belongs to [2, N-1 ]; updating the initial memory state of the long-short time memory network model according to the N-1 th long-short time memory vector, and inputting the N-th characteristic vector into the long-short time memory network model to obtain an output vector; and determining the behavior prediction result of the user to be predicted according to the output vector.
In an exemplary embodiment of the present disclosure, the obtaining behavior data of a user to be predicted in the latest N long periods and converting the behavior data into N feature vectors having a precedence order includes: sequentially counting O kinds of behavior data of a user to be predicted in the latest N long periods according to the short periods, wherein each long period comprises M short periods; and converting the behavior data into N M-O dimensional feature vectors with a sequential order.
In an exemplary embodiment of the present disclosure, the long and short term memory network model at least includes: an input layer comprising M-O neurons for inputting the M-O dimensional feature vectors; a first fully-connected layer comprising P neurons for converting the feature vector into a first intermediate vector of dimension P; the long-time and short-time memory layer comprises Q neurons and is used for converting the first intermediate vector into a Q-dimensional long-time and short-time memory vector according to the initial memory state, and any two of the Q neurons are connected; a second fully-connected layer comprising R neurons for converting the long-short time memory vector into a second intermediate vector of R dimensions; an output layer comprising S neurons for converting the second intermediate vector into an S-dimensional output vector.
In an exemplary embodiment of the present disclosure, the long-short term memory layer is a forward long-short term memory layer, the long-short term memory vector is a forward long-short term memory vector, and the initial memory state is a forward initial memory state; the long-time memory network model further comprises: the backward long-short time memory layer comprises T neurons and is used for converting the forward long-short time memory vectors into backward long-short time memory vectors with T dimensions according to a backward initial memory state, and any two neurons in the T neurons are connected; the second full link layer is configured to convert the backward long-and-short term memory vector into the second intermediate vector; the inputting the first feature vector into a long-short term memory network model to obtain a first long-short term memory vector includes: inputting the first feature vector into a long-short time memory network model to obtain a first forward long-short time memory vector; sequentially, updating the initial memory state of the long-time and short-time memory network model according to the i-1 th long-time memory vector, and inputting the i-th feature vector into the long-time and short-time memory network model to obtain the i-th long-time and short-time memory vector, wherein the step of obtaining the i-th long-time and short-time memory vector comprises the following steps: sequentially, updating the forward initial memory state and the backward initial memory state of the long-short time memory network model according to the i-1 forward long-short time memory vector and the i-1 backward long-short time memory vector respectively, and inputting the i-th characteristic vector into the long-short time memory network model to obtain the i-th forward long-short time memory vector and the i-th backward long-short time memory vector.
In one exemplary embodiment of the present disclosure, S ═ M + 1.
In an exemplary embodiment of the present disclosure, the first and second full connection layers are reli (Rectified Linear Unit) activation layers; the output layer is a Softmax (normalized exponential function) activation layer.
In an exemplary embodiment of the disclosure, the inputting the first feature vector into a long-short term memory network model, and obtaining the first long-short term memory vector includes: inputting the first feature vector into a long-short time memory network model to obtain a first long-short time memory vector and a first output vector;
sequentially, updating the initial memory state of the long-time and short-time memory network model according to the i-1 th long-time memory vector, and inputting the i-th feature vector into the long-time and short-time memory network model to obtain the i-th long-time and short-time memory vector, wherein the step of obtaining the i-th long-time and short-time memory vector comprises the following steps: sequentially updating the initial memory state of the long-short time memory network model according to the i-1 th long-short time memory vector, and inputting the i-th characteristic vector into the long-short time memory network model to obtain the i-th long-short time memory vector and the i-th output vector, wherein i belongs to [2, N-1 ];
the method further comprises the following steps: sequentially determining an ith real vector H according to the ith feature vectoriAnd the ith real vector HiAnd the (i-1) th output vector Gi-1Matching is carried out; and if the following condition (1) is met, judging that the user behavior prediction is invalid at this time:
wherein K is a matching threshold; and if the condition (1) is not met, executing the step of determining the behavior prediction result of the user to be predicted according to the output vector, wherein the output vector is the Nth output vector.
According to an aspect of the present disclosure, there is provided a user behavior prediction apparatus including: the data acquisition module is used for acquiring behavior data of a user to be predicted in N long periods and converting the behavior data into N feature vectors with a sequence; the model analysis module is used for inputting a first feature vector into a long-short time memory network model to obtain a first long-short time memory vector, sequentially updating an initial memory state of the long-short time memory network model according to an i-1 th long-short time memory vector, inputting the i feature vector into the long-short time memory network model to obtain an i long-short time memory vector, updating the initial memory state of the long-short time memory network model according to an N-1 th long-short time memory vector, and inputting an N feature vector into the long-short time memory network model to obtain an output vector, wherein i belongs to [2, N-1 ]; and the result output module is used for determining the behavior prediction result of the user to be predicted according to the output vector.
According to an aspect of the present disclosure, there is provided an electronic device including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the method of any one of the above via execution of the executable instructions.
According to an aspect of the present disclosure, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of any one of the above.
Exemplary embodiments of the present disclosure have the following advantageous effects:
acquiring behavior data of a user to be predicted in the last N long periods, converting the behavior data into characteristic vectors, sequentially processing the characteristic vectors by using a long-time and short-time memory network model, synchronously updating an initial memory state of the long-time and short-time memory network model, finally obtaining an output vector corresponding to the N characteristic vector, and determining a behavior prediction result of the user to be predicted according to the output vector. On one hand, by collecting N long-period behavior data of the user and sequentially analyzing the behavior data according to the time sequence through the long-time memory network model and the short-time memory network model, the long-term characteristics and the change trend of the user behavior can be mined, so that the user behavior can be more accurately predicted; on the other hand, the long-time memory network model is a special neural network model, and can extract the features of the data of the feature vectors from the machine dimension, so that the limitation of manually extracting the features is avoided, and the accuracy of user behavior prediction is improved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and together with the description, serve to explain the principles of the disclosure. It is to be understood that the drawings in the following description are merely exemplary of the disclosure, and that other drawings may be derived from those drawings by one of ordinary skill in the art without the exercise of inventive faculty.
FIG. 1 schematically illustrates a flow chart of a user behavior prediction method in the present exemplary embodiment;
FIG. 2 is a schematic diagram that schematically illustrates an embodiment of an long-and-short term memory network model;
FIG. 3 schematically illustrates a schematic diagram of a long and short term memory layer in the present exemplary embodiment;
FIG. 4 is a schematic diagram that schematically illustrates another long-and-short memory network model in the present exemplary embodiment;
FIG. 5 schematically illustrates a flow chart of another user behavior prediction method in the present exemplary embodiment;
fig. 6 is a block diagram schematically showing the structure of a user behavior prediction apparatus in the present exemplary embodiment;
fig. 7 schematically illustrates an electronic device for implementing the above method in the present exemplary embodiment;
fig. 8 schematically illustrates a computer-readable storage medium for implementing the above-described method in the present exemplary embodiment.
Detailed Description
Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many different forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
An exemplary embodiment of the present disclosure first provides a user behavior prediction method. The user behavior prediction means predicting whether the user makes or does not make a certain behavior within a certain time in the future, or predicting the degree of making a certain behavior, for example, predicting whether the user gets coupons and the number of times of getting in a week in the future, predicting whether the user logs in a client program and the time of logging in three days in the future, and the like.
Referring to fig. 1, the user behavior prediction method may include steps S110 to S150:
step S110, acquiring behavior data of the user to be predicted in the latest N long periods, and converting the behavior data into N feature vectors with a sequence.
In the present exemplary embodiment, the long period may be a time period for predicting user behavior, for example, predicting user behavior in a week in the future, the long period may be a week, and in step S110, the behavior data of the user to be predicted in the last N weeks may be counted. The behavior data of each long period may be converted into a feature vector, for example, various behavior data in each long period may be counted and arranged into the feature vector according to a certain sequence, or the behavior data in each long period may be arranged into the feature vector according to a time sequence. The behavior data of each long period should be converted into eigenvectors with the same form and dimension according to the same rule, and the N eigenvectors can be numbered according to the time sequence of the N long periods, wherein the eigenvector corresponding to the earliest long period is the first eigenvector, and the eigenvector corresponding to the latest long period is the nth eigenvector. N may be any positive integer, and the larger the value of N is, the more sufficient the behavior data of the user to be predicted is obtained, the more favorable the accurate prediction in the following, but the difficulty of obtaining the behavior data and the subsequent computation amount need to be considered, and the above factors may be integrated to set the value of N according to the actual situation.
Step S120, inputting the first feature vector into a long-short time memory network model to obtain a first long-short time memory vector.
The long-short time memory network model is a neural network model adopting a long-short time memory layer as a circulation middle layer. In the exemplary embodiment, the trained long-time memory network model is used for predicting the user behavior. Inputting the characteristic vector into a long-time memory network model to obtain the numerical value of each neuron and an output vector; the values of the long and short term memory layer are arranged according to the sequence of the neurons, and a long and short term memory vector, namely a middle layer vector of the model, can be obtained. The first feature vector corresponds to a first long-short term memory vector, the second feature vector corresponds to a second long-short term memory vector, and the rest is done in sequence.
And S130, sequentially updating the initial memory state of the long-short-time memory network model according to the i-1 th long-short-time memory vector, and inputting the i-th feature vector into the long-short-time memory network model to obtain the i-th long-short-time memory vector, wherein i belongs to [2, N-1 ].
In the exemplary embodiment, the long-short term memory layer of the long-short term memory network model has an initial memory state, and the initial memory state can be regarded as a hidden neuron. The first neuron of the long-time and short-time memory layer is determined by the initial memory state and the previous intermediate layer. Therefore, when the ith feature vector is processed by using the long-term memory network model, the initial memory state needs to be updated in addition to the input of the ith feature vector. Specifically, the initial memory state of the first feature vector may be considered to be 0, or may be artificially set to another value; and then, the initial memory state of the current long-short time memory network model can be determined according to the previous long-short time memory vector, so that the long-short time memory network model is used for sequentially processing each characteristic vector.
Step S140, updating the initial memory state of the long-short time memory network model according to the N-1 th long-short time memory vector, and inputting the N-th feature vector into the long-short time memory network model to obtain an output vector.
And sequentially processing each feature vector by using the long-time and short-time memory network model until the last feature vector is the Nth feature vector. In processing the nth feature vector, the determination method of the initial memory state is the same as that of step S130. After the initial memory state is updated, the nth feature vector is input into the long-short time memory network model, so that an nth long-short time memory vector and an output vector can be obtained.
And S150, determining a behavior prediction result of the user to be predicted according to the output vector.
The dimension of the output vector generally has practical significance and represents the classification result of behavior prediction, for example, whether the user is predicted to receive the coupons in the next week or not is predicted, the output vector can be a two-dimensional vector, the numerical values of the two dimensions can respectively represent the probability that the user is predicted to receive and not receive the coupons, or the number of the user to receive the coupons in the next week, the output vector can be a B + 1-dimensional vector, the numerical value of the first dimension represents the probability that the user is predicted to receive one coupon, the numerical value of the second dimension represents the probability … that the user is predicted to receive two coupons, the numerical value of the second dimension represents the probability that the user is predicted to receive B coupons, and the numerical value of the B +1 dimension represents the probability that the user is predicted not to receive the coupons.
In the present exemplary embodiment, the output vector represents the behavior prediction classification of the (N + 1) th long period (i.e. one long period in the future), so the output vector may be converted into a behavior prediction result, specifically, the dimension with the largest value in the output vector may be used as the behavior prediction result, for example, in the above example of predicting the number of coupons that the user gets in a week in the future, if the value of the dimension C of the output vector is the largest, the prediction result is that the user gets C coupons in the week in the future; in addition, a certain threshold value may also be set, when the maximum value in the output vector reaches the threshold value, the model prediction is considered to be successful, and the behavior prediction result is output, otherwise, the model prediction may be considered to be failed, and more behavior data is needed. For example, when the user is predicted to receive the coupon in the next week, the obtained output vector is [0.6,0.4], which indicates that the probability of the user to receive the coupon in the next week is predicted to be 60% and the probability of not receiving the coupon is predicted to be 40%, and if the threshold value is set to be 70%, although the probability of the user to receive the coupon in the next week is relatively high, the model predicts that the result is less than 70%, the reliability of the prediction is insufficient, and the result of prediction failure can be output. Subsequently, the process returns to step S110, where the value of N is increased, the type of behavior data is increased, and the above steps S120 to S150 are repeated.
Based on the above description, in the present exemplary embodiment, behavior data of the user to be predicted in the last N long periods is obtained, and is converted into the feature vectors, and then the long-time and short-time memory network models are sequentially used to process the behavior data, and the initial memory state of the long-time and short-time memory network models is synchronously updated, so that an output vector corresponding to the nth feature vector is finally obtained, and a behavior prediction result of the user to be predicted is determined according to the output vector. On one hand, by collecting N long-period behavior data of the user and sequentially analyzing the behavior data according to the time sequence through the long-time memory network model and the short-time memory network model, the long-term characteristics and the change trend of the user behavior can be mined, so that the user behavior can be more accurately predicted; on the other hand, the long-time memory network model is a special neural network model, and can extract the features of the data of the feature vectors from the machine dimension, so that the limitation of manually extracting the features is avoided, and the accuracy of user behavior prediction is improved.
In an exemplary embodiment, step S110 may be implemented by:
and sequentially counting O kinds of behavior data of the user to be predicted in N long periods according to the short periods, wherein each long period comprises M short periods.
And converting the behavior data into N M-O dimensional feature vectors with a sequential order.
The short period refers to a basic data statistics period, and corresponds to a long period, for example, the long period may be one week, the short period may be one day, and each long period includes 7 short periods. In the present exemplary embodiment, the long period is an integral multiple of the short period.
When the behavior data are counted, various related behavior data can be counted according to the behavior predicted as required. For example, in a scenario of predicting the user to receive the coupon, taking one day as a short period, the following 5 behavior data may be counted: whether a user signs in on the coupon getting platform every day can be represented by 0/1; the number of times that a user signs in the platform every day; whether the user gets the coupons every day can be represented by 0/1; the number of times the user has a field coupon per day; the total coupon value of the coupons the user receives each day. The 5 kinds of data of each day are counted to obtain 35 data in a long period of one week, so that the behavior data of each long period can be converted into a 35-dimensional feature vector. And counting the behavior data of a plurality of long periods to obtain a plurality of 35-dimensional feature vectors, wherein the feature vectors have a sequence corresponding to the long periods.
M and O may be any positive integer according to the requirement of practical application, and the disclosure is not limited thereto.
In an exemplary embodiment, as shown with reference to fig. 2, the long term memory network model may include at least the following neural network layers:
the input layer comprises M-O neurons and is used for inputting M-O dimensional feature vectors;
a first fully-connected layer comprising P neurons for converting the feature vector into a first intermediate vector of dimension P;
the long-short time memory layer comprises Q neurons and is used for converting the first intermediate vector into a Q-dimensional long-short time memory vector according to an initial memory state, and any two of the Q neurons are connected;
the second full-connection layer comprises R neurons and is used for converting the long-time memory vector into a second intermediate vector of an R dimension;
and an output layer including S neurons for converting the second intermediate vector into an S-dimensional output vector.
The characteristic vectors input by the input layer represent behavior data of M short periods of a user, have certain time sequence characteristics, can be fully extracted by the first full-connection layer, can be mined by the long-time memory layer, can be sorted by the second full-connection layer, and are finally converted into output vectors by the output layer.
In this exemplary embodiment, referring to fig. 3, any two neurons in the Q neurons of the long-short term memory layer are connected, and a certain weight is set to indicate that states can be transmitted between any two or more neurons. In addition, it is also possible to set the state of one-way transmission only in the long-term memory layer, for example, Q neurons are arranged in order as C1、C2…CQThen, for any i, j ∈ [1, Q ]]When i is>j is, there is CiTo CjConnection weight W (C)i->Cj) 0. In particular, it is also possible to provide that the state is transmitted only by the last neuron, i.e. if and only if i ═ j-1, then there is CiTo CjConnection weight W (C)i->Cj) Not equal to 0. The present disclosure is not particularly limited thereto.
In an exemplary embodiment, the number of neurons in the long term memory layer may be on the order of 1000, e.g., Q may be 512, 1024, 1536, etc.
As shown in fig. 2, when each feature vector is processed by using the long-term memory network model, the initial memory state in the next round of model processing can be calculated from the long-term memory vectors obtained by the long-term memory layer, and the initial memory state can participate in the next round of model processing. Therefore, the time sequence characteristics of the behavior data of the previous long period are transmitted to the processing process of the behavior data of the next long period, and the time sequence characteristics of the behavior data of the long period are mined.
It should be noted that, in one user behavior prediction, N rounds of predictions are performed by using the long-term and short-term memory network model, where the long-term and short-term memory network model used in each round of prediction is connected to the long-term and short-term memory network model in the previous round through the long-term and short-term memory layer, and the connection weight may be regarded as a cyclic connection weight of the long-term and short-term memory layer (i.e., a weight for calculating an initial memory state), and for any round of prediction, the weight is a parameter in the model and should be the same.
For the long-time memory network model, a large amount of historical user data can be used for training. For example, behavior data of a plurality of users in N +1 historical long periods are obtained, the earlier N long periods of the behavior data are converted into sample feature vectors, and the 2 nd to N +1 th long periods of the behavior data are converted into behavior labels, so that two corresponding relationships between the feature vectors and the behavior labels are generated: one is that each feature vector corresponds to the behavior tag of the next long period; the other is that each set of feature vectors (containing N feature vectors) corresponds to the behavior label of the (N + 1) th long period. The internal connection weight of the long-term memory network model and the cyclic connection weight of the long-term memory layer and the short-term memory layer can be trained through the corresponding relation between the two feature vectors and the behavior labels (namely, the weight of the initial memory state is calculated).
Further, the long-time and short-time memory layer is a forward long-time and short-time memory layer, the long-time and short-time memory vectors are forward long-time and short-time memory vectors, and the initial memory state is a forward initial memory state; then, referring to fig. 4, the long-term memory network model may further include: the backward long-short time memory layer comprises T neurons and is used for converting the forward long-short time memory vectors into backward long-short time memory vectors with T dimensions according to a backward initial memory state, and any two neurons in the T neurons are connected;
accordingly, a second fully-connected layer may be used to convert the backward long-short term memory vector into a second intermediate vector.
Step S120 may include: and inputting the first feature vector into a long-short time memory network model to obtain a first forward long-short time memory vector and a first backward long-short time memory vector.
Step S130 may include: sequentially, updating the forward initial memory state and the backward initial memory state of the long-short time memory network model according to the i-1 forward long-short time memory vector and the i-1 backward long-short time memory vector respectively, and inputting the i-th characteristic vector into the long-short time memory network model to obtain the i-th forward long-short time memory vector and the i-th backward long-short time memory vector.
By adding a long-short time memory layer, the vectors in the first full-connection layer can be fully mined for time sequence characteristics, and the accuracy of user behavior prediction is further improved. It should be understood that, according to actual needs, a third or more long-term memory layers may also be provided, and other intermediate layers may also be inserted between the long-term memory layers, and a third full-connection layer, a fourth full-connection layer, and the like may also be added, which is not particularly limited in this disclosure.
It should be noted that P, Q, R, S, T represents the number of neurons in different layers in the long-term and short-term memory network model, and the number of neurons is determined by the parameter setting condition of the developer and the training condition of the model, and is specifically related to a plurality of factors such as the application scenario, the complexity of the user behavior data, the number of sample data, and the training parameters. Accordingly, P, Q, R, S, T may be any positive integer, and the disclosure is not limited in this respect.
In an exemplary embodiment, in the long-term memory network model, bias units (which may be considered as special neurons in form) may be disposed in any layer or layers.
In an exemplary embodiment, the long-term memory network model may be used to predict the number of short cycles a user performs a certain action in the next M short cycles, such as the number of days to take a coupon in the next 7 days. The prediction result should be an integer between [0, M ], and the user behavior prediction can be converted into a classification problem of M +1 results, so that S ═ M +1 can be obtained.
In an exemplary embodiment, the first and second fully connected layers may be ReLu activated layers; the output layer may be a Softmax active layer. The ReLu activation function is used in the full connection layer, and sparsification can be carried out while the features are extracted. And the Softmax activation function is suitable for carrying out normalization processing on the output numerical value to obtain a classification result.
In an exemplary embodiment, the user behavior prediction method may be as shown in fig. 5, and includes the following steps:
step S510, counting behavior data of a user to be predicted in the latest N long periods, and converting the behavior data into N feature vectors with a sequence;
step S520, inputting the first feature vector into a long-short time memory network model to obtain a first long-short time memory vector and a first output vector G1;
step S530, sequentially, updating the initial memory state of the long-short time memory network model according to the i-1 th long-short time memory vector, inputting the i-th characteristic vector into the long-short time memory network model, and obtaining the i-th long-short time memory vector and the i-th output vector GiWhere i is e [2, N-1]]；
Step S540, updating the initial memory state of the long-short time memory network model according to the N-1 th long-short time memory vector, and inputting the Nth characteristic vector into the long-short time memory network model to obtain an Nth output vector;
step S550, sequentially determining the ith real vector Hi according to the ith feature vector, and converting the ith real vector H into the ith real vectoriAnd the (i-1) th output vector Gi-1Matching is carried out;
step S560, if the following condition (1) is satisfied, determining that the user behavior prediction is invalid this time:
wherein K is a matching threshold, II (-) is an indication function, and values 1 and 0 are respectively taken when the (-) in the parenthesis is true and false;
and step S570, if the condition (1) is not met, determining a behavior prediction result of the user to be predicted according to the Nth output vector.
In the long-time and short-time memory network model, the output vector represents the user behavior prediction of the next long period, for example, the first output vector G1 corresponding to the first feature vector is the prediction result of the second long period. In the present exemplary embodiment, the first long period to the nth long period are the latest N long periods, that is, history periods, and the real behavior result may be obtained according to the second feature vector of the second long period, and is represented by the real vector H2. Comparing H2 with G1, it can be known whether the prediction result of G1 is accurate. In the present exemplary embodiment, GNG1-G are output vectors that are ultimately needed to predict the user's behavior over a long period of time in the futureN-1The intermediate data generated in the model processing process can be regarded as the intermediate data, and the accuracy of model prediction can be reflected to a certain extent. Thus, H is sequentially matchediAnd Gi-1And cumulatively counting the degree of matching
Characterized, a larger value indicates a higher model accuracy. A matching threshold K can be set as a criterion if
If the effect of the model in the previous N-1 round of prediction is not good, the result of the Nth round of prediction is also not credible, and the fact that the user behavior prediction is invalid at this time can be judged; on the contrary, the prediction effect is better, and can pass through GNAnd determining a final behavior prediction result. The value of K may be set empirically, for example, K is 0.7 · (N-1), which means that the model requires at least 70% of round number prediction accuracy in the previous N-1 round of prediction, or the value of K may be optimized in the actual application process, for example, the valid user behavior prediction result may be compared with the subsequent actual user behavior, if the mismatch is more, the value of K may be increased appropriately, and the invalid user behavior prediction result may be compared with the subsequent actual user behaviorIn many cases, the value of K may be appropriately reduced, and the present disclosure is not particularly limited thereto.
In other embodiments, since the long-time and short-time memory network model has a time sequence cumulative effect, and the larger the number of predicted rounds is, the more reliable the predicted result is, the matching condition of the predicted result of each round can also be weighted and counted, that is, when the following condition (2) is satisfied, it can be determined that the user behavior prediction is invalid this time:
wherein, the weight of each round of prediction can be equal to the number of rounds. K in condition (2) is a different form of matching threshold from K in condition (1), and the values are usually not equal, for example, K in condition (1) is 0.7 · (N-1), and K in condition (2) may be 0.7.
Specifically, which determination condition is adopted, the present disclosure is not particularly limited.
An exemplary embodiment of the present disclosure also provides a user behavior prediction apparatus, as shown in fig. 6, the apparatus 600 may include: the data acquisition module 610 is configured to acquire behavior data of a user to be predicted in N long periods, and convert the behavior data into N feature vectors with a precedence order; the model analysis module 620 is configured to input the first feature vector into a long-short term memory network model to obtain a first long-short term memory vector, sequentially update an initial memory state of the long-short term memory network model according to the i-1 th long-short term memory vector, input the i-th feature vector into the long-short term memory network model to obtain an i-th long-short term memory vector, update the initial memory state of the long-short term memory network model according to the N-1 th long-short term memory vector, and input the N-th feature vector into the long-short term memory network model to obtain an output vector, where i belongs to [2, N-1 ]; and a result output module 630, configured to determine a behavior prediction result of the user to be predicted according to the output vector.
In an exemplary embodiment, the data obtaining module may be further configured to count, according to the short periods, the O kinds of behavior data of the user to be predicted in the last N long periods in turn, where each long period includes M short periods, and convert the behavior data into N M × O dimensional feature vectors having a sequential order.
In an exemplary embodiment, the long and short term memory network model includes at least: the input layer comprises M-O neurons and is used for inputting M-O dimensional feature vectors; a first fully-connected layer comprising P neurons for converting the feature vector into a first intermediate vector of dimension P; the long-short time memory layer comprises Q neurons and is used for converting the first intermediate vector into a Q-dimensional long-short time memory vector according to an initial memory state, and any two of the Q neurons are connected; the second full-connection layer comprises R neurons and is used for converting the long-time memory vector into a second intermediate vector of an R dimension; and an output layer including S neurons for converting the second intermediate vector into an S-dimensional output vector.
In an exemplary embodiment, the long-time and short-time memory layer is a forward long-time and short-time memory layer, the long-time and short-time memory vectors are forward long-time and short-time memory vectors, and the initial memory state is a forward initial memory state; the long-term memory network model may further include: the backward long-short time memory layer comprises T neurons and is used for converting the forward long-short time memory vectors into backward long-short time memory vectors with T dimensions according to a backward initial memory state, and any two neurons in the T neurons are connected; the second full link layer is used for converting the backward long-short time memory vector into a second intermediate vector; the model analysis module can be further used for inputting the first feature vector into a long-short time memory network model to obtain a first forward long-short time memory vector, sequentially updating a forward initial memory state and a backward initial memory state of the long-short time memory network model according to the i-1 forward long-short time memory vector and the i-1 backward long-short time memory vector respectively, and inputting the i-th feature vector into the long-short time memory network model to obtain the i-th forward long-short time memory vector and the i-th backward long-short time memory vector.
In an exemplary embodiment, S ═ M + 1.
In an exemplary embodiment, the first and second fully connected layers may be ReLu activated layers; the output layer may be a Softmax active layer.
In an exemplary embodiment, the model analysis module may be further configured to input the first feature vector into a long-short time memory network model to obtain a first long-short time memory vector and a first output vector, sequentially update an initial memory state of the long-short time memory network model according to an i-1 th long-short time memory vector, and input the ith feature vector into the long-short time memory network model to obtain an ith long-short time memory vector and an ith output vector, where i belongs to [2, N-1]](ii) a The result output module can be further used for sequentially determining the ith real vector H according to the ith feature vectoriAnd the ith real vector HiAnd the (i-1) th output vector Gi-1Matching, if the following condition (1) is met, judging that the user behavior prediction is invalid at the current time, and if the following condition (1) is not met, determining a behavior prediction result of the user to be predicted according to an output vector, wherein the output vector is an Nth output vector;
wherein K is the matching threshold.
The details of the modules are described in detail in the embodiments of the method section, and thus are not described again.
Exemplary embodiments of the present disclosure also provide an electronic device capable of implementing the above method.
As will be appreciated by one skilled in the art, aspects of the present disclosure may be embodied as a system, method or program product. Accordingly, various aspects of the present disclosure may be embodied in the form of: an entirely hardware embodiment, an entirely software embodiment (including firmware, microcode, etc.) or an embodiment combining hardware and software aspects that may all generally be referred to herein as a "circuit," module "or" system.
An electronic device 700 according to such an exemplary embodiment of the present disclosure is described below with reference to fig. 7. The electronic device 700 shown in fig. 7 is only an example and should not bring any limitation to the functions and the scope of use of the embodiments of the present disclosure.
As shown in fig. 7, electronic device 700 is embodied in the form of a general purpose computing device. The components of the electronic device 700 may include, but are not limited to: the at least one processing unit 710, the at least one memory unit 720, a bus 730 connecting different system components (including the memory unit 720 and the processing unit 710), and a display unit 740.
Where the memory unit stores program code, the program code may be executed by the processing unit 710 such that the processing unit 710 performs the steps according to various exemplary embodiments of the present disclosure as described in the above-mentioned "exemplary methods" section of this specification. For example, the processing unit 710 may execute steps S110 to S150 shown in fig. 1, or may execute steps S510 to S570 shown in fig. 5, or the like.
The storage unit 720 may include readable media in the form of volatile memory units, such as a random access memory unit (RAM)721 and/or a cache memory unit 722, and may further include a read only memory unit (ROM) 723.
The memory unit 720 may also include programs/utilities 724 having a set (at least one) of program modules 725, such program modules 725 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.
The electronic device 700 may also communicate with one or more external devices 900 (e.g., keyboard, pointing device, bluetooth device, etc.), with one or more devices that enable a user to interact with the electronic device 700, and/or with any devices (e.g., router, modem, etc.) that enable the electronic device 700 to communicate with one or more other computing devices. Such communication may occur via an input/output (I/O) interface 750. Also, the electronic device 700 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network such as the internet) via the network adapter 760. As shown, the network adapter 760 communicates with the other modules of the electronic device 700 via the bus 730. It should be appreciated that although not shown in the figures, other hardware and/or software modules may be used in conjunction with the electronic device 700, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.
Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein may be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (which may be a CD-ROM, a usb disk, a removable hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which may be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the exemplary embodiments of the present disclosure.
Exemplary embodiments of the present disclosure also provide a computer-readable storage medium having stored thereon a program product capable of implementing the above-described method of the present specification. In some possible embodiments, various aspects of the disclosure may also be implemented in the form of a program product comprising program code for causing a terminal device to perform the steps according to various exemplary embodiments of the disclosure described in the above-mentioned "exemplary methods" section of this specification, when the program product is run on the terminal device.
Referring to fig. 8, a program product 800 for implementing the above method according to an exemplary embodiment of the present disclosure is described, which may employ a portable compact disc read only memory (CD-ROM) and include program code, and may be run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto, and in this document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
A computer readable signal medium may include a propagated data signal with readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Program code for carrying out operations for the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., through the internet using an internet service provider).
Furthermore, the above-described figures are merely schematic illustrations of processes included in methods according to exemplary embodiments of the present disclosure, and are not intended to be limiting. It will be readily understood that the processes shown in the above figures are not intended to indicate or limit the chronological order of the processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, e.g., in multiple modules.
It should be noted that although in the above detailed description several modules or units of the device for action execution are mentioned, such a division is not mandatory. Indeed, the features and functions of two or more modules or units described above may be embodied in one module or unit according to an exemplary embodiment of the present disclosure. Conversely, the features and functions of one module or unit described above may be further divided into embodiments by a plurality of modules or units.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is to be limited only by the terms of the appended claims.
Claims (10)
1. A method for predicting user behavior, comprising:
acquiring behavior data of a user to be predicted in the latest N long periods, and converting the behavior data into N feature vectors with a sequence;
inputting the first feature vector into a long-short time memory network model to obtain a first long-short time memory vector;
sequentially updating the initial memory state of the long-short-time memory network model according to the i-1 th long-short-time memory vector, and inputting the i-th feature vector into the long-short-time memory network model to obtain the i-th long-short-time memory vector, wherein i belongs to [2, N-1 ];
updating the initial memory state of the long-short time memory network model according to the N-1 th long-short time memory vector, and inputting the N-th characteristic vector into the long-short time memory network model to obtain an output vector;
and determining the behavior prediction result of the user to be predicted according to the output vector.
2. The method according to claim 1, wherein the obtaining behavior data of the user to be predicted in the last N long periods and converting the behavior data into N feature vectors with a precedence order comprises:
sequentially counting O kinds of behavior data of a user to be predicted in the latest N long periods according to the short periods, wherein each long period comprises M short periods;
and converting the behavior data into N M-O dimensional feature vectors with a sequential order.
3. The method according to claim 2, wherein the long-term memory network model at least comprises:
an input layer comprising M-O neurons for inputting the M-O dimensional feature vectors;
a first fully-connected layer comprising P neurons for converting the feature vector into a first intermediate vector of dimension P;
the long-time and short-time memory layer comprises Q neurons and is used for converting the first intermediate vector into a Q-dimensional long-time and short-time memory vector according to the initial memory state, and any two of the Q neurons are connected;
a second fully-connected layer comprising R neurons for converting the long-short time memory vector into a second intermediate vector of R dimensions;
an output layer comprising S neurons for converting the second intermediate vector into an S-dimensional output vector.
4. The method of claim 3, wherein the long-short term memory layer is a forward long-short term memory layer, the long-short term memory vectors are forward long-short term memory vectors, and the initial memory state is a forward initial memory state; the long-time memory network model further comprises:
the backward long-short time memory layer comprises T neurons and is used for converting the forward long-short time memory vectors into backward long-short time memory vectors with T dimensions according to a backward initial memory state, and any two neurons in the T neurons are connected;
the second full link layer is configured to convert the backward long-and-short term memory vector into the second intermediate vector;
the inputting the first feature vector into a long-short term memory network model to obtain a first long-short term memory vector includes:
inputting the first feature vector into a long-short time memory network model to obtain a first forward long-short time memory vector;
sequentially, updating the initial memory state of the long-time and short-time memory network model according to the i-1 th long-time memory vector, and inputting the i-th feature vector into the long-time and short-time memory network model to obtain the i-th long-time and short-time memory vector, wherein the step of obtaining the i-th long-time and short-time memory vector comprises the following steps:
sequentially, updating the forward initial memory state and the backward initial memory state of the long-short time memory network model according to the i-1 forward long-short time memory vector and the i-1 backward long-short time memory vector respectively, and inputting the i-th characteristic vector into the long-short time memory network model to obtain the i-th forward long-short time memory vector and the i-th backward long-short time memory vector.
5. A method according to claim 3, wherein S-M + 1.
6. The method of claim 3, wherein the first and second fully connected layers are linear modified unit ReLu activated layers; the output layer is a normalized exponential function Softmax activation layer.
7. The method of claim 1, wherein inputting the first eigenvector into a long-short term memory network model to obtain the first long-short term memory vector comprises:
inputting the first feature vector into a long-short time memory network model to obtain a first long-short time memory vector and a first output vector;
sequentially, updating the initial memory state of the long-time and short-time memory network model according to the i-1 th long-time memory vector, and inputting the i-th feature vector into the long-time and short-time memory network model to obtain the i-th long-time and short-time memory vector, wherein the step of obtaining the i-th long-time and short-time memory vector comprises the following steps:
sequentially updating the initial memory state of the long-short time memory network model according to the i-1 th long-short time memory vector, and inputting the i-th characteristic vector into the long-short time memory network model to obtain the i-th long-short time memory vector and the i-th output vector, wherein i belongs to [2, N-1 ];
the method further comprises the following steps:
sequentially determining an ith real vector H according to the ith feature vectoriAnd the ith real vector HiAnd the (i-1) th output vector Gi-1Matching is carried out;
and if the following condition (1) is met, judging that the user behavior prediction is invalid at this time:
wherein K is a matching threshold;
and if the condition (1) is not met, executing the step of determining the behavior prediction result of the user to be predicted according to the output vector, wherein the output vector is the Nth output vector.
8. A user behavior prediction apparatus, comprising:
the data acquisition module is used for acquiring behavior data of a user to be predicted in N long periods and converting the behavior data into N feature vectors with a sequence;
the model analysis module is used for inputting a first feature vector into a long-short time memory network model to obtain a first long-short time memory vector, sequentially updating an initial memory state of the long-short time memory network model according to an i-1 th long-short time memory vector, inputting the i feature vector into the long-short time memory network model to obtain an i long-short time memory vector, updating the initial memory state of the long-short time memory network model according to an N-1 th long-short time memory vector, and inputting an N feature vector into the long-short time memory network model to obtain an output vector, wherein i belongs to [2, N-1 ];
and the result output module is used for determining the behavior prediction result of the user to be predicted according to the output vector.
9. An electronic device, comprising:
a processor; and
a memory for storing executable instructions of the processor;
wherein the processor is configured to perform the method of any of claims 1-7 via execution of the executable instructions.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811428883.0A CN111221896B (en) | 2018-11-27 | 2018-11-27 | User behavior prediction method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811428883.0A CN111221896B (en) | 2018-11-27 | 2018-11-27 | User behavior prediction method and device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111221896A true CN111221896A (en) | 2020-06-02 |
| CN111221896B CN111221896B (en) | 2025-01-17 |
Family
ID=70830480
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811428883.0A Active CN111221896B (en) | 2018-11-27 | 2018-11-27 | User behavior prediction method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111221896B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112275438A (en) * | 2020-10-13 | 2021-01-29 | 成都智叟智能科技有限公司 | Dry and wet garbage separation and crushing control method and system based on data analysis |
| CN112398663A (en) * | 2020-11-06 | 2021-02-23 | 浪潮云信息技术股份公司 | Elastic IP charging method and system based on deep neural network |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106682220A (en) * | 2017-01-04 | 2017-05-17 | 华南理工大学 | Online traditional Chinese medicine text named entity identifying method based on deep learning |
| CN107730087A (en) * | 2017-09-20 | 2018-02-23 | 平安科技（深圳）有限公司 | Forecast model training method, data monitoring method, device, equipment and medium |
| US10002322B1 (en) * | 2017-04-06 | 2018-06-19 | The Boston Consulting Group, Inc. | Systems and methods for predicting transactions |
| CN108305094A (en) * | 2017-12-18 | 2018-07-20 | 北京三快在线科技有限公司 | A kind of user's behavior prediction method and device, electronic equipment |
| CN108829737A (en) * | 2018-05-21 | 2018-11-16 | 浙江大学 | Text combined crosswise classification method based on two-way shot and long term memory network |
-
2018
- 2018-11-27 CN CN201811428883.0A patent/CN111221896B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106682220A (en) * | 2017-01-04 | 2017-05-17 | 华南理工大学 | Online traditional Chinese medicine text named entity identifying method based on deep learning |
| US10002322B1 (en) * | 2017-04-06 | 2018-06-19 | The Boston Consulting Group, Inc. | Systems and methods for predicting transactions |
| CN107730087A (en) * | 2017-09-20 | 2018-02-23 | 平安科技（深圳）有限公司 | Forecast model training method, data monitoring method, device, equipment and medium |
| CN108305094A (en) * | 2017-12-18 | 2018-07-20 | 北京三快在线科技有限公司 | A kind of user's behavior prediction method and device, electronic equipment |
| CN108829737A (en) * | 2018-05-21 | 2018-11-16 | 浙江大学 | Text combined crosswise classification method based on two-way shot and long term memory network |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112275438A (en) * | 2020-10-13 | 2021-01-29 | 成都智叟智能科技有限公司 | Dry and wet garbage separation and crushing control method and system based on data analysis |
| CN112275438B (en) * | 2020-10-13 | 2022-03-01 | 成都智叟智能科技有限公司 | Control method and system for separation and crushing of wet and dry waste based on data analysis |
| CN112398663A (en) * | 2020-11-06 | 2021-02-23 | 浪潮云信息技术股份公司 | Elastic IP charging method and system based on deep neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111221896B (en) | 2025-01-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110674880B (en) | Network training method, device, medium and electronic equipment for knowledge distillation | |
| AU2020385264B2 (en) | Fusing multimodal data using recurrent neural networks | |
| US20230049747A1 (en) | Training machine learning models using teacher annealing | |
| CN106095942B (en) | Strong variable extracting method and device | |
| CN112202726B (en) | System anomaly detection method based on context sensing | |
| CN109948851A (en) | Method and apparatus for predicting the probability of occurrence of an event | |
| US20200327450A1 (en) | Addressing a loss-metric mismatch with adaptive loss alignment | |
| CN112508265A (en) | Time and activity multi-task prediction method and system for business process management | |
| US12367080B2 (en) | Systems and methods for correlating probability models with non-homogenous time dependencies to generate time-specific data processing predictions | |
| WO2021012263A1 (en) | Systems and methods for end-to-end deep reinforcement learning based coreference resolution | |
| CN114692972B (en) | Training method and device of behavior prediction system | |
| US20240028935A1 (en) | Context-aware prediction and recommendation | |
| CN114298050A (en) | Model training method, entity relation extraction method, device, medium and equipment | |
| CN111221896A (en) | User behavior prediction method and device, electronic device, storage medium | |
| CN116401522A (en) | Financial service dynamic recommendation method and device | |
| CN109272165B (en) | Registration probability estimation method and device, storage medium and electronic equipment | |
| WO2022072772A1 (en) | Causal attention-based multi-stream rnn for computer system metric prediction and influential events identification based on metric and event logs | |
| CN118469381A (en) | Information processing method, apparatus, device, medium, and program product | |
| US12283104B2 (en) | Method, electronic device, and computer program product for extracting target frame | |
| CN117708698A (en) | Class determination method, device, equipment and storage medium | |
| CN117407714A (en) | Method and device for training user behavior characterization model and predicting user behavior | |
| CN113596528B (en) | Training method and device of video push model, server and storage medium | |
| CN115186229A (en) | Method for establishing time sequence prediction model, time sequence prediction method and device | |
| CN111402042A (en) | Data analysis and display method for stock market large disc state analysis | |
| CN114648070B (en) | Critical behavior extraction network, and training method and device of critical attribute extraction network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |