Employee Attrition Prediction Using Classification Models

The term Attrition refers to the voluntary or involuntary discontinuation of employees in an organization. This paper focuses on discussing a systematic flow for predicting Attrition using Data Analysis and Machine Learning techniques. The steps include Data Acquisition, Data Conditioning, Visualization, and Classification by applying the following Classification Models: Support Vector Machine, Decision Tree, K-Nearest Neighbor, Random Forest, and Naive Bayes algorithms in the Python environment. The resulting predictions of classification were evaluated using three performance metrics: Accuracy Score, Confusion Matrix and the ROC Curve. Based on the obtained results, we inferred that Random Forest classifier delivered the highest accuracy being 83.3% whereas Naive Bayes and Support Vector Machine was better in terms of classifying True Positives and indicated greater Area Under Curve values. This paper intends to be of great use to the Organizations aiming at detecting the key causes of Attrition and minimizing them using the power of data.