← Back to Portfolio

Machine Learning • Pricing Analytics • Business Strategy

Predictive Pricing Optimization

Developed and compared seven machine-learning models to predict insurance claim likelihood and cost, identify higher-risk policyholder segments, and support risk-adjusted pricing and deductible recommendations using explainable AI.

Python XGBoost LightGBM SHAP Pricing Analytics
0.7921 Best-performing model ROC-AUC
7 Models Compared across predictive approaches
4 Tiers Policyholder tenure segments evaluated
SHAP Explainable feature importance analysis

Business Challenge

Building smarter insurance pricing using predictive analytics.

Traditional pricing methods often struggle to capture the complex relationships between customer demographics, policy characteristics, and claim behavior. This project explored how machine-learning models could improve insurance pricing accuracy while remaining explainable for business stakeholders.

The Problem

  • Insurance pricing must balance profitability and customer affordability.
  • Traditional statistical models may miss nonlinear relationships.
  • Business leaders need transparent pricing recommendations.
  • High-risk customer segments must be identified accurately.
  • Pricing decisions should reduce risk while remaining competitive.

Project Goal

Compare multiple predictive algorithms, evaluate their performance, explain model decisions using SHAP, and recommend risk-adjusted pricing strategies that improve business decision-making.

Predictive Modeling

Comparing seven machine-learning approaches.

Multiple supervised learning models were trained and evaluated to determine which algorithm provided the best predictive performance while remaining interpretable for business users.

Data Preparation

Cleaned insurance policy data, engineered predictive features, handled missing values, and prepared datasets for machine-learning workflows.

Model Comparison

Evaluated Generalized Linear Models, Random Forest, XGBoost, and LightGBM to compare predictive performance across multiple metrics.

Explainable AI

Applied SHAP analysis to identify the variables that contributed most to claim predictions and pricing recommendations.

Pricing Recommendations

Converted technical model outputs into practical deductible and pricing recommendations for business stakeholders.

Results

Machine learning delivered stronger predictive performance.

Technical Results

  • Compared seven predictive models.
  • Best model achieved ROC-AUC of 0.7921.
  • SHAP improved model transparency.
  • Identified major claim-risk drivers.

Business Impact

  • Improved pricing recommendations.
  • Supported better deductible decisions.
  • Identified higher-risk customer segments.
  • Connected analytics directly to business strategy.

Methodology

A business analytics framework combining predictive modeling, explainable AI, and strategic decision-making.

The project followed an end-to-end analytics process, beginning with business understanding and data preparation, followed by model development, evaluation, explainability, and business recommendation development.

Analytics Process

  • Business understanding
  • Data preparation
  • Feature engineering
  • Predictive modeling
  • Model validation
  • Business interpretation

Tools & Technologies

  • Python
  • XGBoost
  • LightGBM
  • Scikit-learn
  • SHAP
  • Jupyter Notebook

Skills Demonstrated

Technical analytics combined with business strategy.

Machine Learning Business Analytics Predictive Modeling Python XGBoost LightGBM SHAP Data Storytelling Pricing Strategy Risk Analytics

Explore More Projects

Continue exploring Jasmine's analytics portfolio.

View additional business analytics projects, technical case studies, professional experience, and resume.