A hybrid approach of intelligent systems to help predict absenteeism at work in companies

Abstract

In recent years, several surveys have been conducted on absenteeism and how this affects the routine of conducting productive operations in companies. Therefore, having criteria for predicting absenteeism at work can help managers in contingency actions reduce financial losses due to the absence of a worker in their workplace. The objective of this work is to apply the artificial intelligence concepts of a regularized fuzzy neural network, which combines the benefits of artificial neural networks with the fuzzy set theory to obtain more accurate results in predicting corporate absenteeism. The database called absenteeism at work, taken from the UCI Machine Learning Repository, which captured elements of a Brazilian company, was applied in a fuzzy neural network model that allows the calculation of the regressors, defining the estimate of the lack of hours of an employee. The results of the experiments prove that the intelligent model can help in the creation of a specialist system that assists in the prediction of absenteeism.

An Objective-Based Entropy Approach for Interpretable Decision Tree Models in Support of Human Resource Management: The Case of Absenteeism at Work

“…We use a workplace absenteeism dataset for the period between July 2007 and July 2010 from a courier company in Brazil. This 740-sample dataset, which is available at the UCI Machine Learning Repository [ 11 ], has been subject to previous investigations using various machine learning models (see, e.g., in [ 12 , 13 , 14 , 15 ]). Table 1 lists the 21 features of the dataset that reflect work-related and personal factors.…”

“…We now briefly review the main findings of machine learning models that have analyzed absenteeism, and in so doing, we highlight the contributions of the present study. To enable traceability and facilitate comparison with previous research, we use a dataset that was first introduced by [ 11 ] and has been subject to fairly extensive research (see, e.g., in [ 12 , 13 , 14 , 15 ]). Wahid et al [ 12 ], for example, employed various models, such as Decision Tree, Tree Ensemble, Gradient Boosted Tree, and Random Forest, to predict the absence time.…”

“…In [ 13 ], the rate of absence was discretized into just two classes: less than or equal to 5 hours, or greater than 5 hours. Dogruyol and Sekeroglu [ 14 ] and Araujo et al [ 15 ], on the other hand, treated the target variable—absenteeism in hours—as a continuous variable. They analyzed the same dataset using models such as the Backpropagation (BP) Neural Network, the Radial-Basis Function Neural Network (RBFNN), and the Long Short-Term Memory Network (LSTM).…”

Detection of Anomalies in Large-Scale Cyberattacks Using Fuzzy Neural Networks

“…More recent work involves the concepts of robot manipulation and control [110], prediction of chaotic series [103], effort forecast in software building [111], anomalies identification's in children and adolescents locomotion [112], absenteeism at work [113].…”

Fuzzy Rules to Help Predict Rains and Temperatures in a Brazilian Capital State Based on Data Collected from Satellites

“…In addition, in health problem solving, Lim et al [61] used a fuzzy neural network for real-time premature ventricular contraction detection and Wang et al [62] employed the ability of intelligent hybrid models to detect features of cardiac arrhythmia. Finally, in the last decade, scientific work using such techniques has addressed detection of celestial materials (called Pulsars) [63]; prediction of autism in children [64], adolescents [65], and adults [66] through database obtained from mobile devices; breast cancer [67]; absenteeism at work [68]; control for active power filter [69,70]; data knowledge [71]; and speech recognition [72]. The diversity of these models also covers the field of predicting software building efforts [73], cybersecurity [74], help in cryotherapy and immunotherapy treatments [75,76], and different classification and regression problems [77][78][79][80][81][82][83], where the models differ according to the training techniques, fuzzification or defuzzification process, architecture, number of layers, elements present in the model structure, etc.…”