CN120198115A

CN120198115A - A channel merchant automated settlement processing method and system

Info

Publication number: CN120198115A
Application number: CN202510681996.5A
Authority: CN
Inventors: 王存刚; 沈伟锋; 丁宗锐
Original assignee: Hangzhou Jingjia Technology Co ltd
Current assignee: Hangzhou Jingjia Technology Co ltd
Priority date: 2025-05-26
Filing date: 2025-05-26
Publication date: 2025-06-24
Anticipated expiration: 2045-05-26
Also published as: CN120198115B

Abstract

The invention provides a method and a system for automatically settling accounts of channel providers, which relate to the technical field of data processing, wherein the method comprises the steps of obtaining standard sales data; the method comprises the steps of obtaining a rebate rule parameter, configuring a corresponding weight value for each field, carrying out field structure adjustment and weight binding on the rebate rule parameter, obtaining rule items which have a non-empty matching relation with the current sales record as an evaluable set according to each sales record in standard sales data, carrying out field matching score processing on each rule item, screening a target rule item with the highest score, extracting a rebate coefficient of the rule item, carrying out product calculation on the rebate coefficient and the sales number of the corresponding sales record to obtain a rebate amount, carrying out normalization calculation according to the distribution range of the maximum score value and other score values to generate a confidence score, and generating a settlement item set with the rebate amount and the confidence score according to all sales records.

Description

Automatic settlement processing method and system for channel providers

Technical Field

The invention relates to the technical field of data processing, in particular to a method and a system for automatically settling accounts of channel providers.

Background

In the prior art, the settlement processing of the channel merchant is usually carried out by relying on a financial system or an ERP system inside an enterprise. The system gathers according to basic data such as sales orders, invoice and collection records of the channel providers, and generates the corresponding amount of the channel providers through a preset settlement period and a preset calculation rule. In this process, the enterprise typically sets a fixed rebate ratio or rebate condition, and data classification and summarization are accomplished through an automatic script. For example, the system can apply corresponding calculation templates according to the factors such as the class of the channel quotient, the type of the product and the like, and output settlement details and summary reports. After the final result is confirmed by the financial staff, a statement is generated and payment is performed.

In a quaternary sales promotion scenario for home electronics products, a channel dealer often needs to upload its terminal sales data in order for an enterprise to make rebate accounting in accordance with the actual sales volume. Because sales policies often include intersecting rules of product model and time period, for example, "sales of model a product is 8% during 1 month-3 months of 2025, and other times are 5%", existing systems often only support single rule matching. When the data uploaded by the channel dealer contains sales records of a plurality of models and different time periods, the system cannot accurately distinguish the applicable rule ranges, resulting in some rebate calculation errors. For example, the same product sold in 1 month and 4 months may be uniformly applied with a return rate of 5%, so that the accuracy of the calculation is questioned by the channel manufacturer, and the calculation efficiency and the data reliability are reduced due to the need of manual intervention and correction of finance.

Disclosure of Invention

The invention aims to provide a method and a system for automatically settling accounts of a channel and aims to solve the problems in the background technology.

In order to solve the technical problems, the technical scheme of the invention is as follows:

In a first aspect, a method for automated settlement processing by a channel, the method comprising:

Acquiring original sales data, and performing field detection and time sequencing on the original sales data to obtain standard sales data;

Acquiring a rebate rule parameter, configuring corresponding weight values for each field according to field information in the rebate rule parameter, and carrying out field structure adjustment and weight binding on the rebate rule parameter to obtain a weight rule data set, wherein the rebate rule parameter comprises rebate rule numbers, application time ranges, application product models, application channel information and rebate coefficients;

according to each sales record in the standard sales data, acquiring a rule item which has a non-empty matching relation with the current sales record from the weight rule data set as an evaluable set, and performing field matching score processing on each rule item to obtain a rule adaptation score set;

Screening a target rule item with the highest score according to the rule adaptation score set, extracting a rebate coefficient in the target rule item, and performing product calculation on the rebate coefficient and the sales number of the corresponding sales record to obtain rebate amount corresponding to the sales record;

According to the distribution range of the maximum score value and other score values in the rule adaptation score set, carrying out normalization calculation to generate a corresponding confidence score;

and generating a settlement item set with the rebate amount and the confidence score according to all the sales records, and combining the settlement item set to obtain complete settlement detail data.

Preferably, according to the field information in the rebate rule parameter, configuring a corresponding weight value for each field includes:

Extracting the applicable time range, the applicable product model and the applicable channel information field in the rebate rule parameters to form a field information set;

Calculating a field importance index according to the field stability, hit frequency and distribution fluctuation degree of each field in the field information set in the historical sales record;

Setting a weight adjustment factor corresponding to each field according to the field importance index and combining the distribution characteristics of the fields in different product types or sales channels;

performing product processing on the field importance index and the corresponding weight adjustment factor to obtain a field scoring value set;

and carrying out normalization processing on the field grading value set to generate a field weight value set, wherein the field weight value set comprises a time field weight value, a model field weight value and a channel field weight value.

Preferably, performing field structure adjustment and weight binding on the rebate rule parameters to obtain a weight rule data set, including:

According to the field weight value set, matching each field weight value with a corresponding field in the return rule parameters one by one, and carrying out marking enhancement processing on a matching result to generate a weighted field record;

Combining and binding the weighted field records with the rebate rule numbers and rebate coefficients corresponding to the weighted field records to construct unified rule structure entries;

And constructing rule key value combinations for all rule structure entries according to a preset field sequence, and using the rule key value combinations for index identification to generate a weight rule data set.

Preferably, according to each sales record in the standard sales data, a rule item with a matching relationship with the current sales record not being empty is obtained from the weight rule data set as an evaluable set, and field matching score processing is performed on each rule item to obtain a rule adaptation score set, including:

Extracting a time field, a model field and a channel field of each sales record in standard sales data to form a current sales field set;

Performing traversal screening on all rule items in the weight rule data set, identifying rule items with non-empty intersections or fuzzy similar relations between field values and the current sales field set, and generating an evaluable rule set;

for each rule item in the evaluable rule set, calculating the time coincidence degree score of the current sales field set and the time field thereof, the accurate matching score of the model field and the classification similarity score of the channel field respectively;

Performing product processing on the time overlap ratio score, the accurate matching score and the classification similarity score with the time field weight value, the model field weight value and the channel field weight value respectively, and accumulating the product result to obtain a weighted score of the sales record and the rule item;

And combining the weighted scores obtained by calculation of all the rule items into a set, and generating a rule adaptation score set corresponding to the sales record.

Preferably, calculating the field importance index according to the field stability, hit frequency and distribution fluctuation degree of each field in the field information set in the historical sales record comprises:

Based on historical sales record data, carrying out time dimension segmentation statistics on each field in the field information set, and analyzing the value fluctuation range of the field in a continuous period to obtain the stability score of the field;

calculating the frequency proportion of each field value in the sales records based on the historical sales record data to obtain hit frequency scores;

Based on the historical sales record data, carrying out standard deviation or information entropy calculation on the distribution of each field value to obtain the fluctuation score of the field;

And weighting and summing the stability score, the hit frequency score and the volatility score according to a preset proportion to generate a field importance index of the corresponding field.

Preferably, setting a weight adjustment factor corresponding to each field according to the field importance index and combining the distribution characteristics of the fields in different product types or sales channels, includes:

Based on the historical sales record data, extracting a value set of each field in the field information set under different product types, sales areas or channel types to form field-service dimension mapping relation data;

According to the field-service dimension mapping relation data, counting the value distribution density, concentration and distinguishing of each field under different service dimensions to obtain a field distribution characteristic index set;

and according to each index in the field distribution characteristic index set, carrying out preference factor evaluation calculation by combining the importance index of the field, and outputting a weight adjustment factor of each field.

Preferably, for each rule item in the evaluable rule set, calculating a time overlap score of the current sales field set and its time field, an exact match score with the model field, and a classification similarity score with the channel field, respectively, includes:

performing interval intersection calculation on the time field in the current sales field set and the time interval of the rule item, dividing the intersection days by the total days of the rule time interval to obtain a time overlap ratio score;

Carrying out character-level accurate comparison on the product model fields and the rule item model fields in the sales field set, if the product model fields are completely consistent with the rule item model fields, giving full score, and if the product model fields are partially matched with the rule item model fields, carrying out similarity score reduction to obtain accurate matching score;

and performing category mapping or label similarity calculation on the channel fields in the sales field set and the rule item channel fields, and outputting a category similarity score by corresponding the similarity value to the scoring criteria of the channel fields.

In a second aspect, a system for automated settlement processing for a channel, the system comprising:

the sales data acquisition module is used for acquiring original sales data, and performing field detection and time sequencing processing on the original sales data to obtain standard sales data;

the rebate rule management module is used for acquiring rebate rule parameters, configuring corresponding weight values for each field according to field information in the rebate rule parameters, performing field structure adjustment and weight binding on the rebate rule parameters, and generating a weight rule data set;

the rule matching and scoring module is used for acquiring rule items which have a non-empty matching relationship with the current sales records from the weight rule data set as an evaluable set according to each sales record in the standard sales data, and performing field matching scoring processing on each rule item to generate a rule adaptation score set;

the rebate calculation module is used for screening a target rule item with the highest score according to the rule adaptation score set, extracting a rebate coefficient in the target rule item, and carrying out product calculation on the rebate coefficient and the sales number of the corresponding sales records to generate rebate amount;

the confidence coefficient calculation module is used for carrying out normalization calculation according to the distribution range of the maximum score value and other score values in the rule adaptation score set and generating a corresponding confidence score;

And the settlement generation module is used for generating a settlement item set with the rebate amount and the confidence score according to all the sales records, and combining the settlement item set to generate complete settlement detail data.

The scheme of the invention at least comprises the following beneficial effects:

By constructing a multi-field weighted matching mechanism, the problem of inaccurate adaptation caused by the fact that a rebate rule only supports single condition matching in the prior art can be solved. The system builds a multi-dimensional rule adaptation path based on the combination of the time field, the model field and the channel field, and introduces a field matching score and field weight weighting mechanism, so that the matching process of the rebate rule is more refined and logically transparent. Unlike available system with only hit matching processing logic, the present invention can realize the grading of different rules, and the optimal rule is selected for rebate calculation to raise the matching rationality and accuracy obviously.

Further, aiming at the common model-time period crossing rebate strategy in the household appliance sales promotion, the matching strength of the sales records and the rules is quantified through a field overlap ratio scoring and confidence coefficient generating mechanism, so that the optimal rebate rule item corresponding to the sales records can be automatically identified, and the credibility of the matching result can be judged through the confidence score. The scoring model avoids the hidden risk of wrong matching of the rebate amount, and when the confidence score is lower, the marking can be performed in advance or a manual rechecking mechanism can be triggered, so that the settlement error rate and the later correction cost are fundamentally reduced.

In addition, by combining a weight configuration mechanism and a field importance analysis model, the invention can dynamically generate the field weight according to the historical sales behavior data, so that the rebate decision process not only depends on rule setting, but also has data driving capability. Especially when facing the complex rebate policies of different channels and different product types, the system can automatically adjust the field influence, realize the intelligent judgment of 'same rule and multiple view angles', and effectively solve the rebate calculation complexity problem caused by flexible and changeable sales promotion policies. In the whole, the automatic settlement processing scheme provided by the invention can reduce manual intervention, improve the intelligentization level of rebate accounting and the maintainability of the system, and is suitable for the high-efficiency settlement requirement of large-scale channel providers in rebate scenes.

Drawings

Fig. 1 is a flow chart of a method for automatically settling accounts for a channel according to an embodiment of the present invention.

Detailed Description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

As shown in fig. 1, an embodiment of the present invention proposes a method for automated settlement processing of a channel provider, the method comprising:

S100, acquiring original sales data, and performing field detection and time sequencing on the original sales data to obtain standard sales data;

S200, acquiring a rebate rule parameter, configuring corresponding weight values for each field according to field information in the rebate rule parameter, and carrying out field structure adjustment and weight binding on the rebate rule parameter to obtain a weight rule data set, wherein the rebate rule parameter comprises a rebate rule number, an application time range, an application product model, application channel information and a rebate coefficient;

S300, according to each sales record in the standard sales data, acquiring a rule item which has a non-empty matching relation with the current sales record from the weight rule data set as an evaluable set, and performing field matching score processing on each rule item to obtain a rule adaptation score set;

S400, screening a target rule item with the highest score according to the rule adaptation score set, extracting a rebate coefficient in the target rule item, and performing product calculation on the rebate coefficient and the sales number of the corresponding sales record to obtain rebate amount corresponding to the sales record;

S500, performing normalization calculation according to the distribution range of the maximum score value and other score values in the rule adaptation score set, and generating a corresponding confidence score;

And S600, generating a settlement item set with the rebate amount and the confidence score according to all sales records, and combining the settlement item set to obtain complete settlement detail data.

In the embodiment of the invention, by constructing a set of complete automatic settlement processing method for the channel providers, automatic matching and rebate calculation between mass sales data and rebate rules can be realized, the manual auditing strength is obviously reduced, and the accuracy and traceability of rebate calculation are improved. Firstly, by acquiring original sales data and executing field detection and time ordering operations, the integrity and time sequence consistency of all sales records are ensured, and rule matching deviation caused by missing or disorder is avoided. Based on the preprocessed standard sales data, the system introduces rebate rule parameters as settlement basis, including rebate rule numbers, application time ranges, application product models, application channel information and rebate coefficients. By carrying out structural standardization and field weight configuration on the parameters, a weight rule data set which is structured and can be used for evaluation is formed, and a data basis is provided for a subsequent scoring mechanism.

In the actual calculation process, the system selects rule items with matching relation between all field values and the field values from the weight rule data set as an evaluable set aiming at each sales record, so that full traversal is avoided, and the matching efficiency is improved. On the basis, through matching calculation among the time field, the model field and the channel field, weighted summation operation is carried out by combining the weight values of the fields, and a rule adaptation score set between the sales record and each rule item is output. The score set is not only used for screening the target rule item with the highest score as a final settlement basis, but also used for generating a confidence score through score normalization processing and measuring the credibility of the matching result. The confidence score provides an audit reference for the subsequent settlement process, and when the confidence is low, the system or the person can be guided to backtrack and correct the result. Finally, the system combines the rebate amount corresponding to all sales records with the confidence scores to generate a settlement entry set, performs combination operation, outputs standardized settlement detail data, and provides clear-structured data output results for subsequent payment, financial reconciliation or system butt joint.

The method comprises the steps of acquiring original sales data, and carrying out field detection and time sequencing processing on the original sales data, wherein the original sales data is acquired from a sales system uploaded by a channel merchant or in an enterprise, and at least comprises a plurality of fields such as sales time, product model, channel codes, sales number and the like. Considering that the original data has various sources and different formats, the system needs to perform preliminary verification processing on the data integrity and the field consistency. The field detection process is mainly used for identifying the record with the missing field, the null field value or the abnormal data format, and rejecting, repairing or marking the record. After the detection is passed, the system sorts all records in ascending order or descending order according to the sales time field so as to ensure that the time sequence can be accurately quoted in the subsequent rule matching process and avoid cross matching errors. After the operation is finished, the data after the arrangement is stored as standard sales data for subsequent rule adaptation and rebate calculation.

The method specifically comprises the steps of extracting set rebate rule content from a rebate policy management system, wherein the rebate rule parameter comprises fields such as rule numbers, application time ranges, application product models, application channel information, rebate coefficients and the like. These parameters are typically pre-established by the marketing or financial sector and are configured according to the promotion period, product category, or channel level. The system can check the validity period of the rules in the extraction process, so that the expired or non-validated rules are ensured not to be misused in the current calculation flow. In a specific implementation, the system can also establish an index according to the rule number so as to quickly search in a link of matching the sales records with the rules. This step ensures that the rebate calculation logical base data has timeliness and integrity, and can support subsequent field comparison and rebate amount calculation operations.

The method comprises the steps of screening a target rule item with the highest score according to a rule adaptation score set, extracting a rebate coefficient in the target rule item, and carrying out product calculation on the rebate coefficient and the sales number of a corresponding sales record to obtain the rebate amount corresponding to the sales record, wherein the method specifically comprises the following steps: the system sorts the rule adaptation score sets formed by each sales record in the scoring process, and selects the rule item with the highest score value as the optimal adaptation rule corresponding to the record. The rebate coefficient in the target rule item is the rebate proportionality coefficient which should be used by the sales record under the current rule structure. The system multiplies the rebate coefficient with the sales quantity recorded in the sales records, and calculates the rebate amount corresponding to the sales records. For example, in a sales record, if the rebate coefficient of the target rule item is 7% and the sales number is 150, the system calculates the rebate amount as 150 times 7%, which is the rebate value of the record. The step ensures that the calculation of the rebate amount not only is based on the optimal rule option formed by the scoring mechanism, but also maintains the consistency and logic closed loop between the numerical result and the original sales data.

The method specifically comprises the steps of taking the highest score value in a set as a core reference value after the system obtains all rule adaptation scores of a certain sales record, and simultaneously calculating the minimum value of the score values in the set and the span of an integral interval. The system then calculates its confidence level based on the ratio of the location of the maximum in the overall score distribution and outputs a confidence score. The confidence score is used to measure the rationality of the match between the current target rule term and the sales record. For example, when a sales record shows a medium level of match under multiple rule terms and scores are close, the system generated confidence score will be low to suggest that the certainty of the rebate calculation in the subsequent decision flow is weak, and when there is a score significantly higher than the other terms, the confidence score will be close to full score, reflecting a high confidence rule fit result. The confidence scores may be used for marking, grading, or manual intervention referencing in subsequent settlement audit links.

According to all sales records, generating a settlement item set with rebate amount and confidence scores, combining the settlement item set to obtain complete settlement detail data, wherein the settlement item set comprises the following specific steps: the system binds the rebate amount corresponding to each sales record with the confidence score to generate a settlement entry with a structured field. The settlement entry may include field information such as sales record number, rebate amount, matched rule number, confidence score, settlement status, etc. And the system performs statistics summarization operation on all settlement items, classifies and merges the settlement items according to the channel identifier, the product dimension or the period dimension, and outputs final settlement detail data. For example, sales records generated by a channel in a quarter are combined by the system according to sales time to form a rebate for the channel in the quarter. The detail can be directly called by a financial system for payment application, and can also be output as an account checking report, an audit interface or a channel external settlement file. The processing mode not only improves the integrity and output efficiency of the rebate detail structure, but also provides uniform settlement interface support for multidimensional service integration.

In a preferred embodiment of the present invention, configuring a corresponding weight value for each field according to field information in the rebate rule parameter includes:

In the embodiment of the invention, in order to ensure that the fields of the rebate rules have differentiation and weight difference in the matching process, the flexibility and the intelligent degree of rule screening are effectively improved by introducing a field weight configuration mechanism. Specifically, the system extracts fields such as an application time range, an application product model, application channel information and the like from the rebate rule parameters to form a unified field information set. And by combining the historical sales records, the performance characteristics of each field in the actual sales data are counted, so that the field importance index is generated. The analysis content comprises value change frequency of the field in different time periods for evaluating the stability of the field, the frequency of the field referenced in all sales records for judging hit rate, and the distribution breadth and the offset degree of the field value for reflecting the fluctuation. And respectively quantifying the three indexes by using numerical values, setting a preset proportion, and carrying out weighted fusion to form a scoring result reflecting the overall importance degree of the field.

Further, the system takes the value characteristic of the fields under each service dimension as a source of the regulating factors in consideration of the fact that the importance of different fields under different product types or channel strategies is different. And generating weight adjustment factors related to the field matching strategies by extracting the distribution characteristics of the fields under different product classifications and channel scenes, and carrying out product processing on the weight adjustment factors and the field importance indexes. The product results represent the scoring values that the field should have in the current business context, i.e., the set of field scoring values. In order to facilitate the subsequent scoring module to perform weighted computation, the system normalizes all scoring values, ultimately generating a set of field weight values for actual rule matching computation. The processing mechanism avoids the stiffness problem of fixed weight configuration, realizes the coupling of field importance and service situation, and effectively enhances the accuracy of rebate calculation and the intelligence of rule adaptation.

In a preferred embodiment of the present invention, performing field structure adjustment and weight binding on the rebate rule parameters to obtain a weight rule data set, including:

In the embodiment of the invention, aiming at the problems of complex structure and multiple data formats of the rebate rule parameters, a method for adjusting the field structure and binding the weights is provided, so that the rule parameters have uniform data structures and field weight information, and the subsequent scoring and rule calling processes are facilitated. In this embodiment, the system first reads the generated set of field weight values, and matches the weight values in the set with the corresponding time field, model field, and channel field in the rebate rule parameters one by one. After matching is completed, the weight value of each field is added, and marking enhancement processing is carried out, so that the original field not only has original semantics, but also has importance weight information when participating in scoring, and a weighted field record is formed.

And then, the system combines the weighted field records with the rebate rule numbers and rebate coefficients to generate rule structure entries with uniform structures, and each entry simultaneously contains a field value, a field weight and a settlement element and has complete semantic and scoring functions. After all rule item structures are completed, the system combines field values according to a set field sequence (such as time-model-channel) to generate key values for subsequent rule indexes and calls, so that a weight rule data set is formed. The data set has a standardized structure, can efficiently support screening, scoring and decision logic execution in the matching process, and also provides a uniform data interface for system maintenance.

The system specifically comprises the steps of reading the weights of a time field, a product model field and a channel field from the generated field weight value set, and matching the weights to corresponding fields in the rebate rule parameters respectively. For example, if a rebate rule applies to the first quarter of 2025, class a products, and primary distribution channels, the system appends the corresponding weights of the time field, model field, and channel field to the parameter field of the rule so that it is no longer static text, but is a structured field with weight attributes.

In the marking enhancement process, the system performs explicit expression on the relation between the field and the corresponding weight thereof in a structure identification or data label mode, so that the numerical contribution of the field can be correctly identified in the subsequent scoring calculation. For example, the field weights may be encoded into the structure entry as field attachment attributes, or the structure may be defined by a key-value triple of "field name-field value-field weight" to support subsequent weighted matching processes. The enhancement operation ensures that the rebate rule field has complete semantic expression required by participating in scoring calculation, and effectively eliminates the problem of operation interruption or deviation caused by field weight loss in the scoring flow.

The system specifically comprises the steps of structurally combining field values and field weights in each rebate rule with the rebate rule number and the rebate coefficient of the rule item after field marking enhancement is completed, and generating a complete and unified data entry. The rule structure entry is provided with field matching information and rebate calculation parameters, and is a key reference unit in the calculation process of the follow-up sales record score and rebate amount.

To ensure structural consistency and ease of indexing of the entries, the system may arrange the fields in a predefined field order and encapsulate them using a unified data model or structured object. For example, a rule structure entry may be organized in the order of "rule number → time field (with weight) → model field (with weight) → channel field (with weight) → rebate coefficient" and stored in the system in a standard object manner for later recall and comparison. The step ensures the consistency and the integrity of the rebate rule data, and is also beneficial to improving the calculation efficiency and the result traceability of rule matching and evaluation.

The system specifically comprises the steps of connecting or combining field values according to a set field priority order aiming at all generated unified rule structure entries to form a unique rule key value identifier. The key value identifier can be generated by adopting modes such as character string splicing, hash calculation or code mapping and the like, and is used for quickly searching the matching rule in the scoring stage of the sales record, so that the whole rule data set is prevented from being repeatedly traversed.

For example, if the field sequence is set as "time→model→channel", the system sequentially splices the time period identifier, the product model code and the channel category identifier of a certain rule to form a rule key value, such as "2025q1_a001_d1", and then uses the key value as the main index field of the rule structure entry. The system builds a weight rule data set supporting quick query by establishing a mapping relation between the key value and the rule structure item.

The rule data set can establish an index table by key values, and support advanced matching operations such as fuzzy query, prefix matching or similarity calculation. In practical application, the structure remarkably improves the retrieval efficiency in the large-scale rule matching process, is particularly suitable for complex business scenes with hundreds or even thousands of rebate rules, and can effectively reduce the delay and resource consumption in the scoring process.

In a preferred embodiment of the present invention, according to each sales record in the standard sales data, a rule item whose matching relationship with the current sales record is not null is obtained from the weight rule data set as an evaluable set, and field matching score processing is performed on each rule item to obtain a rule adaptation score set, including:

In the embodiment of the invention, in order to improve the matching precision and the grading efficiency between the sales records and the rebate rules, the system can calculate the association strength between the sales records and the rule items in a structural mode by constructing a field grading flow of sub-steps, and provide quantitative support for rebate calculation. For each standard sales data record, the system firstly extracts the time field, the model field and the channel field contained in the standard sales data record to form a current sales field set. The field set serves as an input source for subsequent matching and scoring.

And the system analyzes all rule items in the weight rule data set one by one, and identifies rule items with non-empty intersection, partial coincidence or fuzzy matching relation with the current sales field set by comparing the time, model and channel field value of each rule. A eligible rule term is identified as an evaluable rule set. The definition of the set effectively reduces the consumption of computing resources and avoids the interference of invalid rules on scoring results.

In the scoring phase, the system performs a scoring operation on each of the rules items in the evaluable set separately. The time field is processed in an interval intersection mode, and a time coincidence degree score is output by calculating the overlapping day ratio of the sales time and the rule application time range. The model field uses an accurate matching strategy to compare the character strings of the field character by character, full score is given when the character strings are completely consistent, and score is reduced in proportion according to the similarity degree when the character strings are partially matched. And the channel fields evaluate the generic proximity between the fields through a preset channel label system or mapping rule, and output classification similarity scores. The three scoring values represent how close the sales records are to the rule items in three key dimensions, respectively.

In the score calculation stage, the system performs product operation on the time coincidence degree score, the accurate matching score and the classification similarity score and the field weight value respectively to reflect the duty ratio contribution of each field score in the overall score. All the product results are summed to form a weighted score between the sales record and the current rule term. And after the weighted scores corresponding to all the evaluable rule items are combined, a rule adaptation score set of the sales record is formed. The set not only reflects the matching quality between a plurality of rules and sales data, but also provides an original scoring basis for subsequent rebate calculation and confidence coefficient generation, and enhances the accuracy, transparency and adjustability of the system in the rule adaptation process.

Performing traversal screening on all rule items in the weight rule data set, identifying rule items with non-empty intersections or fuzzy similar relations between field values and the current sales field set, and generating an evaluable rule set, wherein the method specifically comprises the following steps: when processing each standard sales record, the system firstly extracts a time field, a product model field and a channel field from the record to form a current sales field set. The system takes the field set as input and compares the field value with all rule items in the weight rule data set one by one. In the comparison process, the system preferentially judges whether an explicit intersection exists between the current sales field set and the field value of the rule item, namely whether the field value is the same or contained in the application range of the rule item, for example, the time period coincides or the model belongs to the same classification.

If the accurate matching condition is not met, the system further executes fuzzy matching processing, for example, prefix or suffix similarity exists in the product model field, or the channel fields belong to the same channel branch in the category mapping table, and the system can be regarded as that the similarity relationship exists between the fields. In the process, the system sets a matching strategy according to the field types, for example, the system performs interval coincidence judgment on the time field, performs character string similarity comparison on the type number field, and calls channel classification mapping rules on the channel field.

And identifying that at least one field has a valid intersection or a rule item with similarity exceeding a preset threshold in the process, namely adding the rule item into an evaluable rule set of the current sales record. The set is used as an input set of subsequent scoring calculation and rebate selection, so that a scoring mechanism is ensured to focus on rule items with relevance, the calculation efficiency is improved, and the system resource waste caused by participation of irrelevant rules in scoring is reduced.

In a preferred embodiment of the present invention, calculating the field importance index according to the field stability, hit frequency and distribution fluctuation degree of each field in the field information set in the historical sales record includes:

In the embodiment of the invention, in order to improve the scientificity and the distinguishing property of the field weight in the rule matching process, a method for calculating the field importance index based on the historical sales record is introduced into the system. The method performs stability, hit frequency and volatility analysis on the time field, the model field and the channel field extracted from the rebate rule parameters respectively, so as to ensure that the field weight has sufficient historical data support.

In the data analysis stage, the system firstly segments the historical sales records according to time periods, such as statistics of the value change of each field by month or quarter. For each field, the system calculates the amplitude of the value change in the continuous period, wherein small change indicates that the field value is high in stability, and the system assigns the field value to be higher in score. The stability field generally reflects channel characteristics or product sales patterns more reliably and is therefore more valuable as a reference in rebate judgment.

In the hit frequency scoring process, the system counts the frequency of each field value in the history record, and takes the proportion of the frequency to the total record number as a scoring basis. The high frequency description field has a popularity for sales activities and a relatively high score. The system also identifies field values that occur only in a very small number of records to reduce their score and reduce noise effects.

In terms of fluctuation evaluation, the system adopts a standard deviation or information entropy algorithm to analyze the discreteness of field values in sample distribution. The more concentrated and regular the field values, the higher the score, the broader and more discrete the distribution, and the lower the score. The three scores represent the stability, coverage and predictability of the field in the historical data, respectively.

The system performs weighted summation on the stability score, the hit frequency score and the volatility score according to a preset ratio (such as 3:2:1) to generate a field importance index of each field. The index is directly used for the subsequent weight adjustment factor setting link, and lays a data foundation for reasonable generation of the final weight value of the field.

The method comprises the steps of respectively calculating the time coincidence degree scores of a current sales field set and time fields of each rule item in an evaluable rule set, specifically comprising the steps of extracting the sales time fields of sales records by a system and comparing the sales time fields with the applicable time ranges defined in the rule items. If the sales time is completely within the regular effective time period, the system considers the time to be completely matched and assigns a high-level score. If the sales time only partially overlaps the regular time interval, the system sets a scoring grade according to the proportion of the overlapping days in the regular total time period. If the sales time is well before or after the regular time period, the score is lowest or no score.

For example, if a sales record date is 2025, 4,5, and the time range of the rule term is 2025, 4, 1, to 2025, 4, 30, the system determines that it is fully covered, and gives a full score, and if the sales record date is 3, 31, the rule term is adjacent to the rule time period only, but not overlapping, and no score is given. The scoring ensures the basic screening role of the time field in rule adaptation, and is particularly suitable for scenes in which rebate policies are strongly bound with promotion time.

The method comprises the steps of firstly, carrying out character level comparison on a medium-size number field in a sales record and a medium-size number field in a rule item by a system. If there is some character difference but there is structural similarity, for example, the same product series or prefix is consistent, the system gives the next grade score by the set similarity rule.

In a specific implementation, the system can judge whether the product model has a derivative relationship, for example, "A123-B" and "A123-C" are classified as the same-series products, and if the products cannot be completely matched, the matching score of the middle grade can be still given according to the same-series principle. The strategy is suitable for scenes that the product model names are not uniform or the data reported by the sales terminal has simplified processing, and can maintain the accuracy of scoring judgment on the premise of ensuring the matching flexibility.

The system firstly extracts channel field values from sales records and maps the channel field values into standard channel class labels according to a predefined channel classification table. The system compares the label with the label of the channel field in the rule item to judge the similarity degree of the label in the channel hierarchical tree or the classification mapping structure.

If the sales channels and the regular channels belong to the completely consistent labels, for example, the label is 'direct camping primary distribution', the label is the highest grade, if the upper-lower grade relation or the same class (such as 'direct camping channel' and 'direct camping secondary distribution') exists, the higher grade but not the full score is set according to the similar relation grade, and if the label does not belong to the same class, the grade is lower or the grade is regarded as unmatched. According to the method, through the structural processing of the channel labels, flexible adaptation under a complex channel system is supported, and the limitation that the traditional system cannot process the cross-level similarity problem of channels is effectively solved.

The method specifically comprises the steps of performing product processing on a time overlap ratio score, an accurate matching score and a classification similarity score with a time field weight value, a model field weight value and a channel field weight value respectively, and accumulating product results to obtain a weighted score of the sales record and the rule item, wherein the method specifically comprises the following steps: and after the system obtains the grading result of each field, respectively calling the weight value of the corresponding field in the field weight value set, and multiplying each grading value with the corresponding weight value. The product represents the contribution of the field to the overall score.

All field contribution values are summed together after calculation to form a final weighted score between the current sales record and a certain rule term. The score reflects the overall adaptation degree of the current rule item and the sales record, and is a core decision basis for the subsequent screening of the optimal rebate rule item. According to the method, the field weight control is introduced, so that the importance difference of different fields to the scoring model is reflected, the scoring mechanism has adjustability and interpretability, and the unilateral influence of scoring of different dimensions on a final result is avoided.

In a preferred embodiment of the present invention, setting a weight adjustment factor corresponding to each field according to the field importance index and in combination with the distribution characteristics of the fields in different product types or sales channels, includes:

In the embodiment of the invention, in order to further improve the adaptability of the field weight to the actual business strategy, a system introduces a setting mechanism of the weight adjustment factor on the basis of the field importance index. The mechanism comprehensively considers the distribution characteristics of the fields under different service dimensions, dynamically adjusts the field scores, and enhances the scene generalization capability of the scoring model.

The system firstly extracts all the value conditions of each field under different product types, sales areas and channel types from the historical sales records, and carries out business dimension classification analysis on the values. For example, the product model field may be more decision-making in a high-end product family and less influencing in a low-priced promotional product. The system forms distributed characteristic parameters describing the difference of the fields in the service contexts by identifying the occurrence frequency, the distribution concentration and the category distinguishing capability of the field values under different dimensions.

Then, the system executes adjustment analysis on the distribution characteristic parameters, and combines the adjustment analysis with the field importance index. For example, if a field has a high importance score in the history, while also having a high distribution density in the current product line, the system assigns it a higher adjustment factor value. If the field has higher score, but the value is sparse or mixed in a certain business dimension, the system adjusts the score influence downwards, so that the matching deviation is avoided.

The system performs weighting or multiplication operation on the adjustment factors and the field importance indexes to generate weight adjustment factors of each field, and acts the weight adjustment factors on the importance indexes to obtain grading correction values. These correction values form a field scoring value set for further normalization to generate a final weight value. The system scoring strategy is guaranteed to have business awareness, inherits data driving characteristics, reflects business flexibility, and improves practicality and accuracy of the rebate rule adaptation model.

Based on historical sales record data, extracting a value set of each field in a field information set under different product types, sales areas or channel types to form field-service dimension mapping relation data, wherein the method specifically comprises the following steps of: the system firstly cleans and clusters the historical sales records, and extracts specific value records of time fields, product model fields and channel fields under different service attributes (such as product classification, region coding or channel level). The system generalizes the frequencies, the value types and the belongings of the fields under different service dimensions to form the mapping between the fields and the service dimensions.

In a specific implementation, the system takes a product model field as an example, and records specific values of the field in different product lines such as high-end, basic, entry and the like, and counts the coverage proportion of the field under each type. Similarly, channel fields will be mapped into categories of camping, distribution, platform, etc. The data is not directly used for scoring, but is used as a basic data structure for downstream analysis of field distribution characteristics, and is used for revealing the use characteristics of the fields under different service conditions. This mapping not only helps the system understand in which business dimensions a field is active, but also provides data support for field scoring.

According to field-service dimension mapping relation data, counting the value distribution density, concentration and distinguishing of each field under different service dimensions to obtain a field distribution characteristic index set, wherein the system performs three types of statistical analysis on each field from the field-dimension mapping data constructed in the first step.

First, the distribution density reflects whether field values are widely present in each traffic dimension. The system quantifies the degree of coverage of a field by calculating the frequency of occurrence of the field in each dimension and comparing it to the total amount of sales records in that dimension. For example, if a field value is present in multiple product types, the density score will be higher.

Second, concentration is used to evaluate whether field values are concentrated within a certain few specific dimensions. The system analyzes the frequency distribution of the field value in all dimensions and gives a concentration score by judging the specific gravity of the main value concentrated in a few dimensions. The higher the concentration, the more strongly focused the field is on the traffic.

Finally, the distinguishability reflects the discrimination capability of the field value to the service dimension. The system determines whether the field helps to distinguish between different traffic scenarios by analyzing how poorly the field is distributed in each dimension, i.e., the proportional difference between the most frequent and least occurring dimensions of the field. For example, if a channel field frequently appears in a camping channel and rarely appears in a third party platform, its discriminative score is higher. These distribution feature scores will be used as a basis for the subsequent generation of weight adjustment factors.

The system respectively takes the distribution density, the concentration and the distinguishing grading value obtained in the last step for each field, and calculates the weight adjusting factor through a comprehensive grading model by combining the importance index value obtained in the historical sales record of the field. The computational model may be a linear weighted structure, a piecewise evaluation strategy, or a fusion function set according to empirical rules to reflect the actual importance of the field in the current business context.

During comprehensive analysis, the system sets corresponding adjusting weights for each distribution characteristic so as to control the duty ratio of the distribution characteristic in the overall grading. For example, the system can correspondingly increase the weight of the discriminatory scores when the discrimination capability of the fields is more focused in part of the service, and increase the influence proportion of the distribution density scores if the fields are more focused and widely applied to multiple dimensions. Finally, the system combines the three scoring results after adjustment with the importance index of the field to generate the weight adjustment factor of the field. The factor is further applied to field grading values in a weight generation module, so that dynamic adjustment of different service characteristics is realized.

For example, in one implementation scenario, the importance score of the model field is high, but its distribution density in different product types is too low, and the system sets the generated adjustment factor to be slightly lower than the original score, thus suppressing the excessive impact of this field on the scoring model. The strategy realizes the organic integration of history data driving and service strategy guiding, so that the field weight configuration has high flexibility and service adaptability.

In a preferred embodiment of the present invention, the weight adjustment factor has a calculation formula as follows:

;

wherein:

Is a field Weight adjustment factors of (2);

Is a field Is derived from the weighted scores described above;

Scoring the distribution density;

scoring the distribution concentration;

Scoring for distinctiveness;

、、 For the adjustment coefficients corresponding to the three feature scores respectively, the system can set or train and learn;

Is a field The number of the values under all service dimensions is derived from the field and service mapping data;

Is a field In the business dimensionCounting the occurrence times of the occurrence times;

For business dimension The total number of sales records is used for normalization and solving the dimension problem;

Is a constant for the total number of all traffic dimensions, number of dimensions;

Is a field The total occurrence times of all service dimensions are used for calculating the relative frequency of the fields;

ensuring that the concentration degree score is between 0 and 1 for entropy normalization denominator;

And Is a fieldThe maximum and minimum duty cycles in each dimension are used for distinguishing field distribution differences;

The log base numbers not shown in the above calculation formulas are all e, namely natural logarithms are defaults.

In the embodiment of the invention, the system firstly extracts the value condition of each field in different product types, sales areas or channel types based on the historical sales records to form the mapping relation between the fields and the service dimension. On the basis, the distribution density, the distribution concentration and the distinguishing characteristics of the fields in each dimension are calculated respectively. The three types of features are quantized into scoring values through normalization processing respectively. The distribution density measures the extent of the field in each dimension, the concentration reflects whether the field value is concentrated in a few dimensions, and the distinguishability measures the difference of the fields in different dimensions. These statistical features are integrated through formulas and embedded into the field weight adjustment process such that the field scores are no longer determined by static rules, but rather dynamically float based on the data presentation.

By combining the importance index of the field with the expressive power in the service dimension, a multi-level and multi-factor weight generation mechanism is constructed, which not only embodies the data statistics meaning, but also can flexibly adjust the behavior strategy. In the rebate rule matching process, when the importance of a field in a certain channel or product dimension is enhanced, the adjustment factor is correspondingly increased, so that the action weight of the field in the matching score is improved. Otherwise, if the field is too sparse or noisy, its weight will be suppressed, thereby reducing its interference impact in rule screening. The dynamic adjustment strategy can effectively reduce the matching error rate in the complex rebate policy and improve the robustness and adaptability of the scoring model.

Meanwhile, the whole weight adjustment factor generation process supports parameter training and business strategy intervention, can provide an algorithm interface for future system intelligent upgrading, also leaves an algorithm space for iterative optimization of rebate policies, and has remarkable expansibility and industrial application prospect. Compared with the traditional mode of statically configuring field weights based on experience values, the scoring formula has stronger generalization capability and dynamic response capability, and is an innovative computing mechanism with theoretical interpretability and practicality.

Wherein, the 、、The setting method of (2) is as follows:

First, static configuration mode (system settings):

the system administrator or rule designer can manually set the weight ratio of each coefficient according to experience or industry characteristics. For example, if the distribution breadth of a field in each business dimension is emphasized in the enterprise sales policy, the α value may be set to be larger, for example, 0.6, and if the field is emphasized more, the γ value may be set to be 0.5, and other coefficients are reduced appropriately. The mode is suitable for a system scene with low rule change frequency and rapid deployment.

Second, dynamic training mode (model optimization):

The system can utilize target variables (such as settlement accuracy, return error backtracking rate and the like) to establish an optimized target function by introducing the existing return history data as a training sample, reversely adjust the values of alpha, beta and gamma in a scoring model, and iteratively learn the optimal coefficient combination through algorithms such as gradient descent, grid search or Bayesian optimization. The method is suitable for rebate systems with large data volume, complex models and frequent rule change, and can remarkably improve matching quality and automation capability.

Assuming that rebate rule scoring is performed for an enterprise in a "custom channel promotion" business, in this business scenario, channel fields have a strong differentiation between different regional branches, while product models are widely distributed across multiple channels but have limited roles. The system administrator sets parameters based on this experience as follows:

=0.2 (indicating that distribution density is of lower importance);

=0.3 (indicating that concentration effect is general);

=0.5 (indicating maximum differentiating effect);

at this time, if a field has an obvious distinguishing capability under a certain service dimension, even if the coverage of the field is not wide, a larger amplification weight is obtained in the weight adjustment due to a higher gamma value, and the acting force of the field in the overall matching score is finally improved. On the contrary, in the scene of balanced distribution of the whole channel and no obvious boundary distinction, the system can properly lower the gamma value and inhibit the interference of ineffective noise.

The adjustment coefficient mechanism has the following beneficial effects:

Supporting expert knowledge injection and data driving training double modes;

The system can flexibly cope with the field weight difference under different service scenes;

avoiding the problem of structural rigidness or weight bias of the scoring model when the scoring model faces field distribution heterogeneity;

Providing interpretability and customization capability for scoring algorithms.

In a preferred embodiment of the present invention, for each rule item in the evaluable rule set, calculating a time overlap score of the current sales field set and its time field, an exact match score with the model field, and a classification similarity score with the channel field, respectively, includes:

In the embodiment of the invention, in order to promote the interpretability and the calculability of multi-field grading between the sales records and the rule items, the system designs a corresponding grading algorithm for each field, so that the matching process has logic tightness and mathematical basis. The time field, the model field and the channel field are respectively processed in three modes of interval scoring, character comparison and classification similarity evaluation to form a complete field matching scoring system.

For the time field, the system calculates the interval intersection between the sales time of the sales record and the applicable time range of the rule item, counts the number of intersection days, divides the total number of days of the time span of the rule item, and outputs a time overlap degree score between 0 and 1. The score can quantify the coincidence degree between the sales occurrence time and the rule effective time, and the time granularity and the adaptation tolerance are considered.

For the product model field, the system executes character-level accurate comparison, if the model field in the sales record is completely consistent with the rule item, full score is directly given, and if partial characters are matched or prefix/suffix superposition exists, the system calculates a score value by using an editing distance or similarity function, and the score value is reduced according to the similarity. The strategy not only maintains strict matching standard, but also allows fuzzy recognition within a certain range, and adapts to model naming difference.

In channel field processing, the system calls a predefined channel classification table, maps the sales field and the rule field into standard class labels, and generates classification similarity scores based on class hierarchy relations or label similarity indexes. For example, if two fields belong to the same channel class but different subclasses, the system may assign a higher but not full score. All the scoring values are normalized according to the established scoring criteria and then output as the input of the weighted scoring of the next step.

By the scoring mode, the system provides clear, traceable and adjustable matching scoring results for each rule item, and rule screening precision and rebate calculation logic transparency are effectively improved.

The embodiment of the invention also provides a system for automatically settling accounts of the channel providers, which comprises:

It should be noted that, the system is a system corresponding to the above method, and all implementation manners in the above method embodiment are applicable to the embodiment, so that the same technical effects can be achieved.

Embodiments of the present invention also provide a computing device comprising a processor, a memory storing a computer program which, when executed by the processor, performs a method as described above. All the implementation manners in the method embodiment are applicable to the embodiment, and the same technical effect can be achieved.

Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform a method as described above. All the implementation manners in the method embodiment are applicable to the embodiment, and the same technical effect can be achieved.

While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that various modifications and adaptations can be made without departing from the principles of the present invention, and such modifications and adaptations are intended to be comprehended within the scope of the present invention.

Claims

1. A method for automated settlement processing of a channel merchant, the method comprising:

2. The automated settlement processing method of claim 1, wherein configuring the corresponding weight value for each field according to the field information in the rebate rule parameter comprises:

3. The automatic settlement processing method of claim 2, wherein the field structure adjustment and the weight binding are performed on the rebate rule parameters to obtain a weight rule data set, and the method comprises the following steps:

4. The automatic settlement processing method of claim 3, wherein, according to each sales record in the standard sales data, a rule item whose matching relationship with the current sales record is not null is obtained from the weight rule data set as an evaluable set, and field matching score processing is performed on each rule item to obtain a rule adaptation score set, comprising:

5. The automated settlement processing method of claim 2, wherein calculating the field importance index based on the field stability, hit frequency and distribution fluctuation degree of each field in the field information set in the history sales record comprises:

6. The automated settlement processing method of claim 5, wherein setting the weight adjustment factor corresponding to each field according to the field importance index in combination with the distribution characteristics of the fields in different product types or sales channels comprises:

7. The automated settlement processing method of claim 4, wherein calculating the time overlap score of the current sales field set and the time field thereof, the exact match score of the model field, and the classification similarity score of the channel field for each rule item in the evaluable rule set, respectively, comprises:

8. A system for automated settlement processing of a channel, for use in the method of any one of claims 1 to 7, the system comprising:

9. A computing device, comprising:

One or more processors;

storage means for storing one or more programs which when executed by the one or more processors cause the one or more processors to implement the method of any of claims 1 to 7.

10. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a program which, when executed by a processor, implements the method according to any of claims 1 to 7.