Human Resources Sptimization with MultiLayer Perceptron: An Automated Selection Tool

Abstract

This study aims to create an artificial neural network (ANN) model with a multi-layer perceptron (MLP) architecture, designed to analyze the CVs of candidates for the position of sales consultant. To do this, a database of 600 CVs cataloged with scores from 0 to 10 by specialists with experience in recruitment and selection (R&S) is used. Fourteen characteristics are extracted from each CV, including ordinal and nominal attributes. A model with 3 hidden layers is used, which is trained with a split of 80% for training and 20% for testing. The activation function chosen for the hidden layers is the Rectified Linear Unit (ReLU), using the "adam" optimizer with a backpropagation algorithm during training using the Mean Squared Error (MSE) performance metric. The results show that the model is effective, giving a Mean Absolute Error (MAE) of 0.33, MSE of 0.37, Root Mean Squared Error (RMSE) of 0.61, and an r² Score of 0.96. These data not only confirm MLP’s ability to replicate human accuracy but also suggest that such technologies can provide a faster and less biased tool for evaluating CVs. This performance of ANN indicates avenues for future research into the integration of other Artificial Intelligence (AI) technologies to refine the interpretation of less quantifiable characteristics, as well as making the process of R&S of new candidates in companies more agile.