Sales-Forecasting

Dataset not included due to size. Place raw_sales.csv inside data/ folder before running.

Project Overview

This project focuses on forecasting daily revenue for a retail / e-commerce dataset using the ARIMA (AutoRegressive Integrated Moving Average) time series model.

The objective is to analyze historical transaction-level sales data, aggregate daily revenue, and generate short-term revenue forecasts to support business decision-making such as inventory planning and demand estimation.

Dataset Description

The dataset contains transactional sales records with the following key fields:

-> Order ID -> Date -> SKU -> Quantity (Qty) -> Revenue (Amount) -> Shipping City & State -> Category

-> For time series modeling, transaction-level data was aggregated into daily total revenue.

Methodology

-> The forecasting workflow follows a structured time series modeling pipeline:

1. Data Preprocessing

-> Converted Date column to datetime -> Converted Amount to numeric -> Removed missing values -> Sorted data chronologically -> Aggregated transaction data to daily revenue

2. Stationarity Check

-> Performed Augmented Dickey-Fuller (ADF) test -> Applied first-order differencing (d=1) due to non-stationarity

3. Model Implementation

-> ARIMA (1,1,1) model selected -> 80% training data / 20% testing data split -> Model fitted on training dataset

4. Model Evaluation

Model performance was evaluated using:

RMSE : (Root Mean Squared Error) MAE :(Mean Absolute Error) MAPE : (Mean Absolute Percentage Error)

5. Model Performance

Metric	     Value

-> RMSE 118,954.73 -> MAE 81,513.08 -> MAPE 12.98%

Interpretation

A MAPE of 12.98% indicates good forecasting accuracy for retail revenue prediction.