Datainfocom USA, Inc.

Datalnfocom USA, Inc.

Wanli Cheng

Chun Wang

Wensu Wang

Kuikui Gao

Ji Zang

Methods and systems are provided for clustering data for use in training models to predict a loss metric for insurance. The historical data is cleaned and aggregated to a quarterly time granularity and a zip code geographic granularity, and features on which to generate the clusters are selected. Clusters are generated using one or more clustering algorithms, evaluated, and the best set of clusters is selected. Clusters may be grouped together based on cluster characteristics. Each cluster or group of clusters is used to train a development model and a forecast model for the loss metric. The accuracy of the development and forecast models is evaluated, and the clustering process is repeated until the accuracy and stability of the development and forecast models is sufficient. At each iteration, historical data may be removed and/or hyperparameters of the clustering algorithms may be adjusted.

Systems and methods for clustering data to forecast risk and other metrics

Methods and systems are provided for predicting and forecasting loss metrics for insurance. One or more models are created to generate development curves to predict ultimate losses for aggregations of long-tail losses, such as bodily injury claim payouts based on the first few months of payout data and other relevant variables. The relevant variables include internal data about policyholders and claims, and external data. Historical data, including potential influential variables and a target, are used to train a predictive development model. The variables are pre-processed and aggregated to an accident-month granularity, then feature reduction techniques are applied to determine the variables that exert the most influence on the target. Dimensionality reduction techniques are then applied to the remaining variables. The most influential variables and the variables created by dimension reduction are used as the input features to train the development model. One or more additional models are trained to forecast future pure premiums (or a different loss metric) based on ultimate losses predicted by the development model, and other relevant variables.

Growing community of inventors

Houston, TX, United States of America

Wanli Cheng