Everything was developed within <48 hours of the hackathon (see commit history for proof)
- Optimal product pricing suggestion (based on price elasticity)
- Product sales forecast: Note - we originally intended to use
fbprophet
but had an installation issue withpystan
, therefore moved to using ARIMA as a makeshift replacement - User-desired features suggestion (using a BERT-based token classification model)
backend/
: Main service backendfrontend/
: Main service frontendcrawler/
: Product listing & reviews crawler for Amazon/Shopeenlp/
: Labeling tool & model training for product features suggestion model
- Products listing - filtering, using
MobileNet
as image encoder for similarity search - NLP for product reviews analysis
- (potentially) Product sales forecast