Machine Learning Predictive Models for preventing Employee Turnover costs

The problem arises in the turnover of employees in the company, something that represents high costs for the corporation, in the investment in the training of employees, as well as expenses in time and consequences in the dynamics of work in teams, due to the constant change of employees within teams, something that causes instability among employees. By analyzing data quality, doing a thorough profiling of the data and also applying predictive models of Machine Learning algorithms, namely AdaBoost, Decision Tree and Support Vector Machine, the solution (developed in python) allows a better understanding of the reasons behind employee turnover and predicts which employees might leave the company soon, so that the human resources departments can prevent that from happening. The results were accurate for AdaBoost - which revealed an accuracy of 98% - and Decision Tree - which revealed an accuracy of 99% - were appropriate for predicting employee turnover, showing a high precision in the predictions, unlike Support Vector Machine which even though revealed an accuracy of 96%, revealed an Area Under the Curve of 0.50, which means this algorithm was not suitable for this data. This project was a contribution for the HR Predictive Analytics area, which allows companies to prevent employee turnover with contractual or moral incentives.