Chi Square Selector Enhanced Fuzzy Clustering Method for Employee Attrition Prediction
This proposed study aims to predict the presence or absence of attrition among employees by developing Chi-Square selection enabled fuzzy hierarchical clustering model. The Employees Attrition Dataset is collected from IBM open data which is available in Kaggle Repository. This dataset is comprised of 34 attributes with 1470 instances. The raw dataset attribute values are in different ranges, min-max normalization is applied to bring all the attributes fall under a same range. Among 34 attributes, the most potential features are determined by applying a Chi-Square selection algorithm. By finding the relationship of the variables by constructing hypothesis and testing, it determines the significant attributes. With the reduced feature subset, clustering is done byemploying Fuzzy C Means algorithm. The resulting clusters are again clustered in a fuzzy hierarchal manner to overcome the inconsistency among the instances with attrition nature. The simulation result proves the efficacy of the proposed model in the prediction of presence or absence of attrition among employees. To enhance the accuracy of prediction Gaussian Mixture Model is also introduced.