An Optimized Harris Hawks Algorithm for Enhancing ANN Performance in Prediction Tasks Applied in Sales Domain

Sales prediction, the process of predicting future sales performance based on past information, offers valuable information to businesses, allowing them to make informed and profitable decisions. This can be done using machine learning, specifically Artificial Neural Networks (ANNs) which have proven to be effective for prediction tasks. A requisite process before using ANNs as prediction models is its training which is commonly done using back-propagation algorithms. Recently, it has been noticed that swarm intelligence algorithms are better training algorithms for ANNs than backpropagation. In this paper, we propose to employ a high-performing swarm intelligence algorithm, Harris Hawks Optimization (HHO) for the training of ANNs applied to the prediction of iPhone sales in Malaysia. Moreover, we propose an optimized Harris Hawks algorithm and apply it for training the same prediction models. Both the original and optimized HHO algorithms are found to be able to better train the prediction models compared to the usual backpropagation algorithm. Furthermore, the proposed optimized HHO algorithm is found to have higher effectiveness than the original HHO as the ANNs trained by the optimized HHO can achieve a root mean square error of 89.6789 during the training as compared to 112.2448 by those trained using the original HHO.