Predicción de la demanda de llamadas entrantes en un centro de contacto mediante el uso de técnicas de aprendizaje automático

Abstract

The prediction of incoming calls in a contact center is a tool used to anticipate the demand that will be received in a given period, by accurately forecasting, it is possible to plan the necessary resources to meet it properly, offering a quality service and avoiding both shortages and overstaffing, which could generate financial losses and an unsatisfactory experience for customers. For this reason, in the present study, an analysis of historical data of incoming calls in a contact center was carried out with the objective of identifying patterns and trends in demand; for this purpose, advanced data modeling techniques were used, including machine learning and time series models. For the modeling, regression models were proposed to predict the number of incoming calls to agents by day and time range. We then compared the performance of various machine learning models using multiple evaluation metrics, and finally selected the Gradient Boosting Regressor (GBR) proving to have the best performance of the MAPE (mean percentage error) evaluation parameter with a result of 23.35%. Day and time range were found to be the most significant contributors to the prediction. These results are expected to aid contact center demand planning, enabling effective resource management, optimizing operations, reducing costs and improving customer experience.