BWKI × Hackathon Prototype
Trendoro
For a student competition, our team built a demand-forecasting prototype using historical Walmart sales data and a PyTorch LSTM. We explored how those forecasts could inform supply-chain planning.
Back to all projects
Project scope
We developed Trendoro as a student prototype to test a forecasting approach on historical sales data. It was not deployed as a commercial planning tool.
Project focus
Our goal was to turn historical sales into forecasts that planners could inspect and compare with recorded demand.
We focused on the data pipeline, forecasting model and interpretable visualizations rather than building a complete ordering system.
Data and model
The prototype used a Walmart sales dataset from the United States, spanning multiple stores, product categories, and time periods.
Data was prepared with Pandas, scaled for training, and modeled with a PyTorch LSTM network to forecast future demand based on historical sequences.
Evaluation and visualization
Model quality was assessed by splitting the dataset into training and test sets and comparing predictions against actual demand values.
Forecasts were visualized in Power BI to make trends and deviations easier to interpret in a decision-making context.
What remained unfinished
The prototype requires prepared data and a technical setup to run.
A possible next iteration would add an interface for uploading sales data and viewing forecasts without working directly in code.
What we demonstrated
Prototype built under student project and hackathon constraints.
Used historical Walmart sales data across stores, years, and categories.
Visualized forecast outputs to explore demand-planning use cases.