Assignment 3: Building Time Series Forecasts

A data science project focused on analyzing and forecasting retail sales data from Corporación Favorita stores using time series decomposition techniques.

This repository contains a Jupyter notebook that guides you through temporal data analysis to:

• Identify trends, seasonality, and anomalies in retail sales data
• Build forecasting models using time series decomposition
• Predict future sales patterns for specific stores and product families

Repository Structure

Assignment-3/
├── README.md                    # This file
├── starter_notebook.ipynb       # Main analysis notebook with step-by-step instructions
└── data/                        # Data directory (CSV files not included in repo. Download from Kaggle)
    ├── train.csv               # Historical sales training data
    ├── holidays_events.csv     # Holiday and events information
    └── store.csv                # Store metadata

Getting Started

Prerequisites

• Python 3.7 or higher
• Jupyter Notebook or JupyterLab
• Required Python libraries (see Installation section)

Installation

1. Clone this repository:

   git clone <repository-url>
   cd Assignment-3

2. Download the dataset:

   • Go to the Store Sales - Time Series Forecasting dataset on Kaggle
   • Click on the "Data" tab
   • Scroll down and select Download All
   • Extract the zip file and move the following files to the data/ directory:
     • train.csv
     • store.csv
     • holidays_events.csv

3. Install required libraries:

   pip install pandas numpy matplotlib seaborn scikit-learn statsmodels

Running the Notebook

1. Launch Jupyter:

   jupyter notebook

   or

   jupyter lab

2. Open the notebook:

   • Navigate to starter_notebook.ipynb
   • Click to open

3. Execute cells sequentially:

   • Run each cell in order from top to bottom
   • Follow the instructions and complete the TODO sections
   • Checkpoints throughout the notebook help verify your progress