Willis Towers Watson has developed a layered machine learning system for data processing, utilizing decision trees with varying depths and a gradient boosting method. The system iteratively trains to model interaction effects among different depths. The patented system aims to enhance predictive accuracy in insurance premium outcomes. GlobalData’s report on Willis Towers Watson gives a 360-degree view of the company including its patenting strategy. Buy the report here.
According to GlobalData’s company profile on Willis Towers Watson, was a key innovation area identified from patents. Willis Towers Watson's grant share as of February 2024 was 46%. Grant share is based on the ratio of number of grants to total number of patents.
Layered machine learning system for processing insurance premium data
A recently granted patent (Publication Number: US11853906B1) discloses a system and method for generating a layered machine learning model to predict insurance-related outcomes based on historical customer data. The system includes a processor, a database storing a training dataset, and memory storing instructions for model training. The model involves decision trees with different depths and parameters such as weight variables, hyperparameters, and starting values. The training process refines model parameters through iterations, involving computations of derivatives, determining splits, and updating model parameters. The trained model satisfies stopping criteria and stores decision tree parameters for predictions on insurance premium policies, claim cost, frequency, and severity based on customer input data.
Furthermore, the system and method involve computing gain values for decision trees, removing leaf nodes based on certain parameters, and utilizing probability distributions like Gaussian, Poisson, gamma, Tweedie, and logistic. The stopping criteria include a maximum number of iterations, threshold values for additional gain, and performance evaluation. The model is designed to handle various insurance-related predictions and is configured to process data efficiently by converting records to numeric representation, selecting appropriate parameters, and refining model parameters through iterative training. The patent also covers a non-transitory computer-readable medium with processor-executable instructions for implementing the layered machine learning model, emphasizing the importance of accurate predictions in the insurance domain and the systematic approach to model training and evaluation outlined in the claims.
To know more about GlobalData’s detailed insights on Willis Towers Watson, buy the report here.
Data Insights
From
The gold standard of business intelligence.
Blending expert knowledge with cutting-edge technology, GlobalData’s unrivalled proprietary data will enable you to decode what’s happening in your market. You can make better informed decisions and gain a future-proof advantage over your competitors.