Predictive modeling for order quantities in the automotive industry: a statistical and machine learning approach

Date

2024-9-05

Author

Sarışen, Miray

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

105
views

120
downloads

Cite This

The automotive industry is undergoing a shift in focus from a product-centric approach to a customer-centric approach. As a result, after-sales services have become a crucial aspect of providing excellent customer service and a significant source of profit for automotive companies, in addition to traditional manufacturing. To effectively provide these services and products to clients, a well-structured and organized logistics chain is necessary. One example of after-sales services is the delivery of spare parts. In order to meet customer demand, companies must have efficient warehousing and inventory management systems in place. Orders placed by customers for future manufacturing and shipping, but currently unavailable or out of stock, can lead to increased manufacturing costs and dissatisfaction among customers. The goal of this thesis is to use statistical and machine learning methods to improve customer satisfaction, reduce costs and enhance efficiency in the after-sales services of the automotive industry by predicting order quantity in advance. In this study, both traditional statistical methods; ARIMA, ETS, TBATS, and Prophet, and machine learning algorithms; including Random Forest, Gradient Boosting, LightGBM, Long Short-Term Memory (LSTM) networks and Feedforward Neural Networks (FFNN) were employed. The analysis is divided into two parts as univariate and multivariate analysis since the initial purpose of the analysis is to predict the order quantity. Initial analysis is applied on the univariate data, and to improve the accuracy of the predictions, further variables are included. The analysis results showed that for univariate analysis Prophet, and for the multivariate analysis, Random Forest slightly outperformed other models in terms of model evaluation metrics.

Subject Keywords

Prediction, Order quantity, Machine learning, Statistical modeling, Prophet

URI

https://hdl.handle.net/11511/111558

Collections

Graduate School of Natural and Applied Sciences, Thesis