This project explores the application of data analytics to minimize waste in the fast fashion industry. By leveraging machine learning and multimodal analysis, the project aims to optimize inventory management, reduce overstocking, and contribute to more sustainable practices.
The project utilizes the Visuelle 2.0 dataset, incorporating sales information, restock data, and product images to create a comprehensive model for product classification and restock prediction.
Key highlights of the project include:
- Multimodal Analysis: A novel approach combining product images and sales data to classify products into low, medium, and high-selling categories.
- Machine Learning Models: Several machine learning models, including neural networks, were trained and fine-tuned to predict restock quantities and sales performance.
- Time Series Analysis: Time series models like ARIMA were used to analyze restocking patterns and forecast future restock needs.
- GPU Acceleration: Model training was optimized using GPU acceleration with CUDA and Keras-Tuner.