METHODS OF APPLYING MACHINE LEARNING IN USER BEHAVIOR ANALYTICS FOR SAAS: FROM CLASSIFICATION TO PREDICTION
Keywords:
Machine Learning, Churn Prediction, Feature Engineering, artificial intelligence, Behavioral Data, SaaSAbstract
This paper explores the application of machine learning methods for user behavior analytics in SaaS platforms to improve classification accuracy and churn prediction. It addresses challenges such as class imbalance, high data dimensionality, and temporal dynamics. The study compares the effectiveness of traditional and modern models (AdaBoost, transformers), and examines the impact of data preprocessing and class balancing (ADASYN) on performance. In the practical case, significant improvements were achieved after vectorizing categorical features and applying class balancing. The work highlights the potential of ML approaches to enhance retention strategies and improve the effectiveness of SaaS products.
References
Chakraborty, A., Raturi, V., & Harsola, S. (2022). BBE-LSWCM: A Bootstrapped Ensemble of Long and Short Window Clickstream Models. arXiv preprint arXiv:2203.16155. https://arxiv.org/abs/2203.16155
Canay, O., & Kocabicak, U. (2025). Predictive modeling and anomaly detection in large-scale web portals through the CAWAL framework. arXiv preprint arXiv:2502.00413. https://arxiv.org/abs/2502.00413
Yuan, J., Qiu, X., Wu, J., Guo, J., Li, W., & Wang, Y.-G. (2024). Integrating behavior analysis with machine learning to predict online learning performance: A scientometric review and empirical study. arXiv preprint arXiv:2406.11847. https://arxiv.org/abs/2502.00413
Sarker, I. H., Colman, A., Han, J., Khan, A. I., Abushark, Y. B., & Salah, K. (2019). BehavDT: A Behavioral Decision Tree Learning to Build User-Centric Context-Aware Predictive Model. arXiv preprint arXiv:2001.00621. https://arxiv.org/abs/2001.00621
