Title: Modelling the demand trend for automobile parts using machine learning methods

Authors: Oguzhan Akan; Abhishek Verma; Sonika Sharma

Addresses: Department of Computer Science, California State University, Northridge, CA, USA ' Department of Computer Science, California State University, Northridge, CA, USA ' Department of Commerce, Shaheed Bhagat Singh College, University of Delhi, Delhi, India

Abstract: This research utilises the automotive spare parts sales dataset, containing 5,000 spare parts. Being able to accurately model demand trend for spare parts helps in reducing warehousing costs, production costs, and could result in higher sales via fulfilling sales orders in a timely manner. In order to model the demand for automobile spare parts we build machine learning models using K-nearest neighbour, support vector regressor, random forest, gradient boosting, back propagation neural networks, and deep neural networks. We conclude that deep neural networks performs better than the rest of the machine learning models on the spare parts dataset for demand trend modelling with a RMSE validation score of 6.25. After a thorough literature review, it is observed that there are few previous researches done on gradient boosting regression for demand trend modelling. Our research proves that gradient boosting's performance is better than other traditional machine learning models such as K-nearest neighbour, random forest, and support vector regressor.

Keywords: demand modelling; gradient boosting; K-nearest neighbours; KNN; random forest; deep neural networks; DNN; support vector machines; SVM.

DOI: 10.1504/IJAISC.2025.148119

International Journal of Artificial Intelligence and Soft Computing, 2025 Vol.9 No.1, pp.62 - 83

Received: 17 May 2024
Accepted: 27 Jun 2025
Published online: 25 Aug 2025 *