YASA-1 E-Commerce Data Analysis and Forecasting

Introduction

This project is a comprehensive analytical solution developed for YASA-1 LLC, aimed at optimizing business operations and enhancing decision-making capabilities. Tasked by the Smart Business Analytics team, the project focuses on three key areas: demand forecasting, seller and product analysis, and product review sentiment analysis.

Task 1: Demand Forecasting

Objective

Implement a robust demand forecasting system for a short-term period (14 days) starting 7 days from the last date in the data for all product groups, including new products with minimal historical data.

Approaches

1. Machine Learning Forecasting: Utilized regression models, such as Random Forest and Gradient Boosting, to predict future demand.
2. Classical Time Series Forecasting: Employed ARIMA models for time series forecasting.

Task 2: Analysis of Sellers and Products

Objective

Conduct an in-depth analysis of sellers and products on the marketplace, focusing on turnover, identifying sales leaders/outsiders in each area, and exploring the dependence of product weight on turnover and price. Additionally, provide segmentation of sellers and products with actionable business insights.

Analysis Conducted

1. Sellers:
   • Identified sellers with the highest and lowest turnover.
   • Determined leaders and outsiders in sales within each area.
2. Products:
   • Conducted turnover analytics to identify top-performing products.
   • Analyzed the best-selling products in each category.
   • Investigated the relationship between product weight, turnover, and price.
   • Segmented sellers and products to derive meaningful business insights.

Task 3: Analysis of Product Semantics

Objective

Develop functionality to classify product review comments as positive, negative, or neutral, and analyze the correlation between text comments and numerical ratings (1-5). Identify products with the best/worst reviews, and highlight sellers who predominantly receive negative feedback. Additionally, extract and highlight price mentions in comments for competitor price analysis.

Steps

1. Sentiment Analysis:
   • Implemented a classifier to categorize review comments into positive, negative, or neutral.
   • Analyzed the correlation between numerical ratings and text comments.
2. Review Analytics:
   • Identified products with the best and worst reviews.
   • Highlighted sellers who received mostly negative feedback.
3. Price Extraction:
   • Extracted price mentions from review comments.
   • Compared mentioned prices with actual product prices.

Result

A visual representation of the analysis results provided in an report1.pbix file with accompanying code that generated the insights.

Project Structure

• data/: Contains the datasets used for analysis.
• notebooks/: Jupyter notebooks with the analysis and visualizations.
• README.md: Project overview and detailed explanation.