Comparative Analysis of Machine Learning Algorithms on Comprehensive and Cluster-Specific Data in the Auto Insurance Industry

Sammanfattning

In recent years, businesses have been focusing on Customer Lifetime Value (CLV) to achieve better customer relationships and to identify high-value customers for more customized marketing strategies. This thesis contributes by comparing the performance of different machine learning models on cluster-specific data points and the complete dataset from the auto insurance industry. In addition, the study also discovers the most valuable customer cluster and devises customer retention strategies based on significant features that influence CLV. For further empirical analysis, we have selected Principal Component Analysis (PCA) and k-means Clustering for customer segmentation. We have also used Random Forest, XGBoost, and Neural Networks, to predict CLV on comprehensive and cluster-specific data. Applied feature importance and hyperparameter tuning have been used for further insights. Overall, the findings suggest the best performance among the models is by Random Forest and its R^2 improved by 27% while RMSE dropped by 39% after applying the models to every cluster for predicting CLV. For future research, the findings from this study can also be adopted in other insurance industries to see how using clustering techniques helps improve the machine learning models’ performances.