Prediction of Employee Attrition and Analyzing Reasons: Using Multi-layer Perceptron in Spark

Abstract

Employees are one of the important assets of any organization. Sudden and unplanned departures of important employees make a big loss in productivity and failure to meet deadlines of project, cost of hiring for replacement. We describe a framework of prediction model built using Multilayer Perceptron implemented in spark that predicts if particular employees will departure along with the analysis to find reasons and hidden patterns using python correlation graphs. The main retention of this paper is to use the real time dataset collect from the IBM company to find the patterns using data mining tools and predict using machine learning algorithm and predict attrition value for a given employee and the corresponding reasons like poor environment satisfaction level, less salary, poor job involvement, fail to work life balance etc. these predictions will help organization to introduce new schemes and rules promoting productivity and avoiding financial and knowledge loss.