Sampling-Based Categorization of Employee Turnover in IBM HR Analytics

Abstract

In this paper, we analyse the reasons for employee turnover using data from IBM's attrition survey. To begin, we used the correlation matrix to eliminate characteristics that were not strongly linked to others. Second, we found that a worker's monthly income, age, and the number of companies they've worked for all had significant effects on turnover, and we discovered this using Random Forest. Then, we used K-means Clustering to divide the population into two groups. The final quantitative analysis we performed used binary logistic regression, and it revealed that frequent travellers were 2.4 times more likely to abandon the group than infrequent ones. In addition, we discovered that human resources employees are more likely to resign than employees in any other division.