Sales Forecasting Using Random Forest & XGBoost

This project develops a machine learning pipeline to predict daily sales (in units) for each product at each store using a Random Forest Regressor. The goal is to enable accurate demand forecasting, supporting better inventory planning and business growth.

---

Dataset

• Source: Kaggle - Toy Sales Dataset
• The dataset includes:
  • Daily sales per product and store
  • Product and store metadata
  • Cost and price info
  • Inventory records

---

Objective

Predict the number of units sold on a daily basis for each product-store combination using historical trends and behavioral patterns.

---

Features Used

Feature	Description
Day_of_Week, Month	Temporal information
Is_Weekend	Binary flag for weekends
Lag_1_Day, Lag_7_Day	Past sales (1 and 7 days ago)
Rolling_Avg_7	7-day average sales
Sales_Drop_Flag	Flag for recent sales drop
Consecutive_Zero_Sales	Number of consecutive days with zero sales
Category_Avg_Sales	Avg. sales for the product's category
Store_Percentile	Store sales percentile (relative ranking)

---

Evaluation Metrics

• MAE: Mean Absolute Error
• RMSE: Root Mean Squared Error
• R² Score: Proportion of variance explained
• Naive Baseline: Compared to using Lag_1_Day as prediction

How to Run

1. Clone this repo.

2. Install dependencies:

   pip install -r requirements.txt