Time Series Analysis Of Queue Lengths In Call Centres: A Predictive Approach

Efficient queue management remains a critical challenge in modern call centres, where fluctuating call volumes and limited staffing often lead to increased customer wait times, reduced satisfaction, and operational inefficiencies. This study focuses on predicting future queue lengths using advanced time series analysis to enhance service delivery and optimize resource allocation. The primary objective is to develop a robust predictive model capable of forecasting short-term queue behaviour based on historical data patterns. The research utilizes both real and simulated call centre datasets comprising variables such as call arrival rates, service rates, and queue lengths recorded over fixed intervals. Exploratory data analysis and statistical tests for stationarity guide the model selection process. Three primary forecasting techniques—AutoRegressive Integrated Moving Average (ARIMA), Seasonal ARIMA (SARIMA), and Long Short-Term Memory (LSTM) neural networks—are implemented and compared based on accuracy metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The findings reveal that while ARIMA performs well for linear trends, SARIMA handles seasonality more effectively. However, the LSTM model demonstrates superior predictive accuracy in capturing complex, non-linear patterns in the queue length data. The study concludes that predictive time series modelling significantly improves decision-making in call centre operations by enabling preemptive adjustments to staffing and reducing average wait times. These insights hold practical value for call centre managers seeking data-driven solutions for capacity planning and customer service enhancement.