The Relationships Between the Binary Correlation and Accuracy in Logistic Regression for Predicting Employee Attrition

This research paper examines the accuracy of the logistic regression method in predicting employee attrition. Employee attrition is of particular interest to human resources departments due to its impact on organizational performance and employee satisfaction. Specifically, this study aims to establish the relationships between various control variables and the binary response variable representing employee attrition. Using the popular IBM HR Analytics Employee Dataset, we assess the contribution of each control variable to prediction accuracy. Additionally, we determine the correlation between each control variable and employee attrition data. For binary control variables, we use the phi correlation, while for non-binary control variables, we employ the point-biserial correlation coefficient. Our results reveal a positive correlation between the value of the correlation coefficient and the improvement in prediction accuracy, highlighting an important principle in building logistic regression models for similar cases.