This demand forecasting model uses Random Forest regression to predict the units sold of a retail chain (on a weekly basis) given the retail chain's past sales data. RandomizedSearchCV is used to tune the hyperparameters of this model.
Sales data input:
- record_ID
- week
- store_ID
- sku_ID
- total_price
- base_price
- is_featured_sku
- is_display_sku
Output:
- Mean Absolute Error
- week-by-week comparison of predicted values versus actual values
Features included in training model:
- discount
- discount percentage
- lagged features (sales over past 4 weeks)
- moving average of units sold
The dataset that this model was trained on is found at https://www.kaggle.com/datasets/aswathrao/demand-forecasting?resource=download&select=train_0irEZ2H.csv