easyRF : An easy way to perform random forest analysis.

This R package was designed for translational and research data. So, a standard format has been established to insure datasets compatiblity. All datasets must contain these features :

• primary ID : a numeric feature with unique ID for each observation (=rows)
• patient ID : a numeric feature
• timepoint : a numeric feature (if not a longitudinal study, create a timepoint feature and fill with "0"). This feature is not used by easyXgboost
• stimulation : a categorical feature (if stimulation feature is not required, create a stimulation feature and fill with "unstim").
• cohort : a numeric feature. This feature is not used by easyXgboost as cross validation is perform independently.
• outcome : the outcome features that must be studied for the dataset
• valid : a logical feature which specify the observations to include ("1") or to exclude ("0")

1/ First, load the library library("easyRF")

2/ Select your dataset (csv/xls format) and generates the 2 configuration files : rf_metadata.csv, rf_grid_parameters.csv rf_initialize() Now, you just have to complete both configuration files (/"project folder"/easyRF/attachments/).

3/ Perform the random forest analysis run_analysis()

This step will :

• Import parameters : recall parameters saved in configuration files.
• Prepare dataset : clean dataset, convert binary features into compatible format (levels : "1","2"), One-Hot-Encoding categorical features, impute data.
• random forest analysis : perform multiple cross-validated rf analysis and autotuning of the model
• Report : print pdf report of rf models