Olist Brazil - Sales Insights and Predictions

Project Overview:

This project focuses on the analysis of sales data from Olist, a Brazilian e-commerce company. The main objectives include developing insightful dashboards and applying machine learning models to predict customer satisfaction.

Business Understanding:

Olist connects sellers with online marketplaces in Brazil. The challenge was to extract actionable insights from sales data to drive growth and improve customer retention.

Key Objectives:

ETL Process: Data cleaning, transformation, and loading using Talend.

Dashboard Development: Visualize sales performance using Power BI. Machine Learning: Predict customer satisfaction using algorithms such as KNN, Decision Tree, SVM, and XGBoost.

Methodology:

Data Preprocessing: Cleaning and feature engineering were performed on the sales dataset. Missing values were handled, and categorical variables were encoded.

UML Fact table :

Talend: For ETL and data transformation.

Key Technologies:

Visualization:

The Power BI dashboard provides a comprehensive view of product performance, top-selling products, and customer satisfaction metrics.

Power BI: For dashboard creation and data visualization.

Machine Learning Models: Multiple models were built and evaluated using confusion matrix and classification reports, with XGBoost delivering the best results.

Machine Learning Models: XGBoost, SVM, Decision Tree, KNN.

Results:

XGBoost achieved the highest performance in terms of accuracy and precision, making it the best model for predicting customer satisfaction.

Conclusion:

This project provides valuable insights into Olist's sales performance and customer satisfaction, enabling data-driven decisions for marketing and sales strategy.