Modifying activities based on probabilistic weather scenarios

Abstract

Disclosed is technology for modifying activities based on probabilistic weather scenarios. A method can include: receiving weather scenarios and respective probabilities, determining event predictions for each weather scenario using a machine learning model trained with historical weather condition data and historical data about event conditions, clustering the event predictions with a time-series clustering approach to identify similarities amongst the weather scenarios and combine similar event conditions to generate a set of weather analytical scenarios, generating a probability of an occurrence of each weather analytical scenario, the probability being based on a total percentage probability of the weather scenarios corresponding to each respective weather analytical scenario, and determining an activity modification based on at least one aggregated event prediction from the probabilities of the weather analytical scenarios or an aggregate impact of the activity modification over a collection of event predictions corresponding to the weather analytical scenarios and the respective probabilities.