Personalized Recommendations & Demand Forecasting
A hybrid ML system pairing personalized product recommendations with inventory-aware demand forecasting, for cold-start and long-tail problems in e-commerce.
Problem
Online stores have two related problems: showing each customer the right product, and stocking enough of it before the demand arrives. Most off-the-shelf systems handle these separately. Recommenders cold-start badly on new customers, and forecasts ignore the personalization signal entirely.
Approach
I built a hybrid pipeline. It combines content-based and collaborative filtering for the recommendations, then feeds the aggregate demand signal into a forecasting model per SKU. SQLite holds the rolling order history, pandas does the feature engineering and scikit-learn trains the models. A Streamlit interface lets a non-technical operator browse predictions by user, by product and by category.
Result
It works end to end, and one system serves both merchandising and operations. Handling cold-start in the recommender also means the forecast gets cleaner signals.
Get in touch
Have a research idea, a role, or a question about one of these projects?
- Email mawangot2000@gmail.com The best way to reach me.
- LinkedIn tatenda-mawango For work and roles.(opens in a new tab)
- GitHub tatendamawango The code behind these projects.(opens in a new tab)
There's also a contact form.