halpert3/complaint-content-classification-nlp

Leverages lemmatization and TF-IDF vectorization to process 162,400 CFPB consumer complaint narratives, consolidating nine financial product classes into five balanced categories. Implements Multinomial Naive Bayes and Gradient Boosting models achieving 86% macro recall, with an API integration layer enabling real-time classification of up-to-date complaint data. Addresses class imbalance through parameter tuning rather than SMOTE, prioritizing recall to minimize false negatives in complaint routing workflows.

No commits in the last 6 months.