Visual Patterns in an Interactive App for Analysis Based on Control Charts and SHAP Values

The aim of this paper is to describe the tool for combining the information of control charts, widely used for statistical quality control with the information obtained from the process of explanation the classification decision made using the machine learning model. Control charts are used to monitor production and show deviations from normal behavior, especially when upper and lower control limits of measured parameters are defined. Analysis of SHAP explanations allowed us to show some dependencies of classification process. Our application, which allows the user to compare the information taken from both control charts and SHAP plots, helps the user discover anomalies, especially dependencies between exceeding the limits of controlled parameters and the relevance of these values in SHAP calculations.