jingpeicomp/product-category-predict

Implements hierarchical three-level product categorization across 962 categories using jieba Chinese tokenization augmented with a 300K-term domain dictionary, then applies TF-IDF weighting and HashingTF dimensionality reduction before Naive Bayes classification. Integrates with Spark 2.2.1 for distributed model training on large datasets (12M+ products trained in ~10 minutes on 8-core/8GB resources) and provides batch prediction via RESTful endpoints, achieving 82% accuracy on held-out test data using only product titles as input features.

157 stars. No commits in the last 6 months.