Automation of Call Center Workforce Management Using NeuralProphet

This research addresses some of the current as well as widely used agent headcount prediction or forecasting models in the call center systems. Performance provided by the models used today are reasonably good, however there are some drawbacks which can be fixed with the help of AI based time series algorithms. In this paper, four such algorithms namely Holt-Winter's, ARIMA, SARIMA and NeuralProphet were used. For this purpose, the data of hourly call arrivals for a period of 3 years was collected and training and testing was performed on it. The accuracy of the models was compared on the basis of the root mean squared error (RMSE) test performed on them. The proposed work benefits Small Organizations who do not have the money or the resources to afford a separate planning team for effective headcount planning and can only afford limited number of agents in the call center system. The number of calls provided by the model helps to enhance the accuracy of staffing based on call arrival patterns (duration and time of calls), average call handling time and seasonality-time of year (holidays and end of the year calls).