Use of K-Means Clustering and Artificial Neutral Network for Predicting Waiting time of Queue system in different Distributions of Arrival and Service times

Abstract

The combination of Machine learning techniques, K-means clustering and Artificial Neural Network(ANN) for analyzing data has not been fully explored especially for queue systems hence the keen interest in this area of research. Machine learning process selected for analysis plays a major role in the effectiveness of the analysis. This study proposes the use of the K-Means clustering and Artificial Neural Network for developing a structure when making predictions of waiting time of a queue system with different distributions of arrival and service times. This is done to understand data in unique ways. In this study, data that consists of the customers’ arrival and service duration (for exponential and erlang distributions) are simulated using Google Colab platform and thereafter used for the analysis. Data are grouped by distribution with the use of K-Means Clustering Method. It consists of 6 features altogether, with 3 features used for the prediction of waiting time. For the learning of the data, the optimization techniques; Stochastic Gradient Descent (SGD), Adaptive Moment Estimate (ADAM) and an optimization technique (Ajewole et al 2024a) were used for training the data with the Artificial Neural Network (ANN). The model was trained for 400 epochs with 80data used for training and 20(combination of K-means clustering and ANN) adapted for the analysis. The results obtained showed the performance of the new structure as effective for queue system analysis with different distributions. With this structure, making predictions for new arrivals into the queue system will be less time consuming.