Leveraging Artificial Intelligence for Customer Segmentation and Demand Forecasting in the Car Rental Industry

Abstract

The dynamic car rental industry faces significant challenges in demand forecasting, with about 50% of companies reporting inaccuracies that result in fleet utilization rates of only 70-75% instead of the optimal 85-90%. The study integrates customer segmentation and demand forecasting into a framework using various ML models. This study utilized historical rental data from Secured Wheels Car Rental reports in Lagos and Ibadan, Nigeria. The data underwent thorough preprocessing, including cleaning, selecting relevant features, and splitting it for analysis. The study employs decision trees, random forests, and clustering algorithms such as DBSCAN, Agglomerative clustering, Fuzzy-C-Means, and Affinity Propagation for segmentation. To enhance demand forecasting in the car rental industry, key customer segmentation features such as inactivity period, number of reservations, and cluster groups were incorporated into the model. This integration allowed for more precise demand predictions by capturing segment-specific patterns. For demand forecasting, the study uses ARIMA, regression model, and Holt-Winters. Performance metrics like accuracy, precision, silhouette coefficient, and Mean Absolute Error (MAE) evaluated the models, and the framework's results were benchmarked against existing methods. Results indicate that the Agglomerative clustering achieved a silhouette coefficient of 0.9238 and a Davies-Bouldin index of 0.0031. At the same time, the HW model recorded a lower Mean Absolute Error (MAE) of 29.3641 and a Mean Squared Error (MSE) of 1183. The HW model was trained with customer segmentation features and the five cluster groups. These enhanced blended models enable more tailored marketing strategies and personalized customer experiences, increasing customer satisfaction and loyalty.