Predição de desligamento de funcionários por meio de modelos de regressão logística

Abstract

The objective of this study is to verify through logistic regression which factors may in uence the increase of the turnover rate of a company's employees. The data comes from a company that makes telephone sales in Uberlândia-MG. We analyzed 264 telesales operators of both sexes and ages between 18 and 48 years old and also 18 behavioral and performance variables of the sellers that represent possible causes that can impact for the increase of turnover. Of all the variables evaluated, ve were signi cant for the prediction of turnover ("Proportion of goals hit", "Number of times the customer goal was beaten", "Quantity of returns in the evaluated period", "Number of absences in the period However, the variable "PropMetaBatida"has a signi cant impact on the reduction of the variable, so that if there is an increase in motivating actions that impact the target, the corresponding variable will have a positive expectation and consequently reduction in turnover. Reviewing the variable remuneration weights according to the real importance for the business is also presented as a measure of low turnover. To select the most parsimonious model, the Backward criterion was used with the Akaike statistic and to evaluate the quality of t, the Hosmer-Lemeshow test was performed.