Abstract: 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.

Abstract: A system includes at least one server implementing a weather forecast engine and in communication with a network, the server to access satellite imagery, published weather predictions, and local measured data via the network; the weather forecast engine to train regional modules using the satellite imagery, published weather predictions, and local measured data; and the weather forecast engine to apply the satellite imagery, published weather predictions, and local measured data to the trained regional modules to obtain regional forecasts. A method for forecasting a weather indicator includes receiving satellite imagery; processing the satellite imagery to generate a weather feature set; applying the weather feature set to a regional module of a weather forecast engine; and forecasting the weather indicator with the weather forecast engine.

Abstract: A system for merging data into a database is disclosed. During operation, the system may fetch a set of data from a data source external to the database. The system may determine that the fetched set of data is unstructured data, and then transform the fetched set of data into structured data. The system may also determine one or more lowest denominators for the fetched set of data, determine that the fetched set of data does not meet the one or more lowest denominators, and transform the fetched set of data to meet the one or more lowest denominators. The system may further determine one or more joinable keys for the fetched set of data, and merge the fetched set of data into the database.