Data Science
Level: Intermediate
Machine Learning for Insurance
การประยุกต์ใช้การเรียนรู้ของเครื่องในธุรกิจประกันภัยและวิทยาศาสตร์ข้อมูล
Bridging traditional actuarial science with modern data science: XGBoost, LightGBM, neural networks, fraud detection, customer churn, and interpretable ML.
Gradient Boosted Decision Trees Objective
Course Registration & Inquiry
Available as individual mentorship, group cohort, or in-house corporate training tailored to your team's dataset.
Lecturer Yodharin Monplub
Format Online / On-site / Hybrid
Prerequisites Basic Statistics / Excel
Comprehensive Course Syllabus & Learning Modules
1
1. Overview of Machine Learning in Insurance
- Comparison between GLM vs Machine Learning in an actuarial setting
- Supervised vs unsupervised learning workflows in insurance operations
- The end-to-end model development lifecycle: from data pipeline to production API
2
2. Data Preparation for Insurance Modeling
- Handling high-cardinality categorical features (Vehicle make/model, postal code, ICD-10 codes)
- Imbalanced classification techniques (SMOTE, Focal Loss, threshold tuning)
- Preventing data leakage and handling temporal splits in insurance datasets
3
3. Tree-Based Ensemble Models & Regularization
- Random Forests, Gradient Boosting Machines, XGBoost, LightGBM, and CatBoost
- Poisson and Gamma deviance loss functions for insurance loss cost prediction
- Hyperparameter optimization using Optuna and cross-validation
4
4. Applications in Life & Health Insurance
- Predictive underwriting and accelerated policy approval
- Predicting lapse, surrender, and policy persistency with survival analysis
- Hospitalization claim cost forecasting and chronic disease progression models
5
5. Applications in Non-Life & Claims Management
- Telematics driving score modelling for usage-based auto insurance (UBI)
- Automated fraud detection using anomaly detection and graph algorithms
- Estimating ultimate claim severity at first notice of loss (FNOL)
6
6. Model Explainability & Governance (XAI)
- SHAP (Shapley Additive exPlanations) values for individual prediction explanations
- Partial Dependence Plots (PDP) and Accumulated Local Effects (ALE)
- Compliance with insurance regulation, fairness, transparency, and bias audits
7
7. Workshops (การฝึกปฏิบัติ)
- Python workshop: Motor frequency-severity prediction with LightGBM and Tweedie regression
- Building a claims fraud detection model with SHAP explanations dashboard