Prediction of Claims Litigation Propensity
InsuranceBusiness Impacts
97%
Model Accuracy
85%
Precision in identifying cases that will result in litigation
86%
Reduction in features through feature engineering
Customer Key Facts
- Country : America
- Size : 1000-5000
- Industry : Insurance
- About : The customer is a large commercial P&C insurer and a subsidiary of a fortune top 20 companies
Challenges
- Difficulty in analyzing trends and insights in the available dataset
- Difficulty in identifying attributes from the data to successfully predict the propensity of litigation
- Difficulty in developing a sophisticated AI-based system for claims litigation propensity prediction
Technologies Used
Colab Notebooks
Compute Engine
Cloud Storage
Power BI
Solution
Quantiphi team prepared a sophisticated AI-based claims propensity prediction algorithm with the following steps
- Thorough data cleaning to ensure that the data is fit for modelling.
- Performed feature engineering; including feature creation and attribute manipulation to ensure the most important features are selected
- Extensive ML experimentation to ensure that best model is selected for propensity prediction
- Model Explanation to showcase model decisioning
Results
High Model Accuracy
- Achieved 97% model accuracy in predicting litigation propensity.
- The model demonstrated 85% precision in identifying cases likely to result in litigation
Efficient Feature Selection
- 86% Reduction in Features: Streamlined the feature set through effective feature engineering, improving model performance and efficiency.
