Product Classification API Part 1: Data Acquisition

Eugene Yan · 9 years ago · 44

A developer built a product classification API that categorizes products based on their titles using Amazon product metadata. Starting with 9.4 million products, the dataset was cleaned by removing incomplete records, excluding certain categories like Books and Movies, filtering to deepest categories only, and requiring minimum 10 products per category, resulting in 4.59 million usable products across 15,000 categories. The cleaned dataset enables training a machine learning model to automatically assign products to their most relevant categories from thousands of options.