/alchemy-broker-modeling

Performed segmentation analysis and predictive modeling on insurance broker performance to conclude a random forest model (highest AUC of 73%) predicted whether 2020 Gross Written Premium will increase or decrease from 2019 with a misclassification rate of 35%. Four classification models (classification trees, logistic regression, random forests, and support vector machines) were built, evaluated, and then tuned for prescriptive measures to analyze broker performance. Explored, visualized, and described five groups of brokers using principal component analysis.

Primary LanguageR

Watchers