Automatic Classification Customer Goods

Project information

This repository presents a comprehensive feasibility study and implementation of an automatic product classification engine for an e-commerce marketplace. It investigates whether product categories can be reliably inferred from textual descriptions and images using both unsupervised clustering and supervised deep learning approaches. The project covers exploratory data analysis, multimodal feature engineering (classical and deep methods), dimensionality reduction, clustering validation, CNN-based image classification with data augmentation, hyperparameter optimization, external data collection via API, and deployment considerations. The objective is to assess the robustness and scalability of automated category assignment to replace manual labeling in a production environment.