Implementation of Machine Learning Algorithms to Predicting Customer Churn for HRMS Software Vendors

Abstract

Introduction: In this competitive market it is very difficult to retain the customers, so it is very important for the HRMS software vendors to predict the customer churn to build the efficient customer retention strategies to work proactively while maintaining the profitability in the business.

Objectives: The objective of this research paper is to build a framework for customer churn prediction from the context of HRMS software vendors using the machine learning algorithms.

Methods: A customer data-set of proprietary HRMS software vendor is used to experiment various machine learning algorithms likes Decision Tree, Random Forest, Logistic Regression, Light GBM, XGBoost, two stack ensemble models are used one is Decision Tree as base model and Random Forest as meta model & another is Logistic Regression as base model and Support Vector Classifier as meta model. The dataset is pre-processed, categorical values are converted to numeric by label encoding technique, for class imbalance issue SMOTE technique is used and domain specific features are selected.Sentiment Analysis is used to read ‘description’ and ‘solution’ columns and analysed the sentiments in 1 & 0. Further to this feature engineering is performed and created the target variable from the dataset.

Results: As a result, stack ensemble model i,e Decision Tree as base model and Random Forest as meta model achieved high accuracy as compared to other machine learning models used individually.

Conclusions: This study provides a promising stack ensemble model to predict the customer churn for HRMS software vendors. This study provides a very practical and proactive approach to HRMS software vendors to build an efficient strategy to retain the customers. In future work explainable AI or deep learning techniques can be used to interpret the machine learning           models working and to improve model performance