SQL-Driven-Business-Insights-with-Machine-Learning is a comprehensive business data analysis and machine learning project that demonstrates how structured SQL analysis and Python-based machine learning can be combined to generate meaningful business insights from raw sales data.
The project is designed to replicate a real-world analytics workflow where data is first explored and aggregated using SQL and then further analyzed and modeled using Python. The primary focus areas include seller performance evaluation, repeat customer behavior analysis, and product category trend identification.
Modern businesses generate large volumes of transactional data, but raw data alone does not provide value unless it is properly analyzed. This project aims to show how SQL can be used for efficient data extraction and aggregation, while Python and machine learning can be used to interpret patterns, classify performance levels, and support strategic decision-making.
- Identify top-performing sellers based on revenue
- Analyze seller location and performance levels
- Measure repeat customer ratio to evaluate customer loyalty
- Analyze product categories to determine high and low performers
- Apply machine learning to classify business metrics into interpretable levels
- Build an end-to-end SQL to machine learning pipeline
├── business_data_analysis.sql
├── ml_business_insights.py
├── data/
│ └── csv datasets
├── requirements.txt
└── README.md
The project uses structured sales data containing information related to sellers, customers, orders, and product categories. The datasets are assumed to be clean and stored in relational tables or CSV files that can be queried using SQL and loaded into Python for further analysis.
Key data entities include:
- Sellers and their geographic information
- Customers and order counts
- Product categories and revenue values
The SQL file is responsible for extracting, aggregating, and preparing the data required for further analysis and machine learning.
The SQL code calculates total revenue generated by each seller by aggregating order-level sales data. Grouping is performed on seller identifiers along with city and state to provide geographic context.
This step allows identification of sellers who contribute the most to overall revenue and helps classify seller performance based on revenue thresholds.
Using conditional logic such as CASE statements, sellers are categorized into different performance levels based on their total revenue. These levels help in segmenting sellers into high, medium, or low performers.
This classification simplifies downstream analysis and supports business decisions such as incentive allocation and seller prioritization.
The SQL code identifies repeat customers by counting the number of orders per unique customer. Customers with more than one order are treated as repeat customers.
The total number of repeat customers is compared with the total customer count to compute the repeat customer ratio. This metric is crucial for understanding customer loyalty and long-term business sustainability.
Revenue is aggregated at the product category level to understand which categories contribute the most to total sales. Categories are ranked based on revenue, enabling identification of best-performing and underperforming product segments.
This information is later used in Python to assign trend levels and perform further analysis.
The Python script builds on the output generated by SQL and applies data processing and machine learning techniques to derive deeper insights.
The script loads CSV outputs or query results into Pandas DataFrames. Data cleaning steps include handling missing values, type conversion, and sorting or filtering records as required.
This ensures the data is suitable for analysis and modeling.
Relevant features such as total revenue, order counts, and aggregated metrics are selected or derived to be used as inputs for machine learning models.
Feature scaling or transformation may be applied where necessary to improve model performance and interpretability.
Machine learning models are used to classify sellers into predefined performance levels based on revenue and other features. Simple and interpretable models are preferred to ensure that results are understandable to non-technical stakeholders.
The model output helps validate the rule-based classifications performed in SQL and provides consistency in performance evaluation.
Although repeat customer ratio is primarily computed using SQL, Python is used to validate results, perform additional analysis, and prepare metrics for reporting or visualization.
This step demonstrates how SQL and Python can complement each other in an analytics workflow.
Product categories are classified into trend levels based on revenue distribution. Categories with higher revenue are assigned higher trend levels, indicating strong market demand.
This classification can support inventory planning, marketing strategy, and category-level decision-making.
- Supervised or rule-based classification techniques are used
- Emphasis is placed on simplicity and interpretability
- Models are designed to support business insights rather than complex prediction tasks
- Results are aligned with SQL-based aggregations for consistency
Run the SQL file in your preferred database system to generate aggregated results:
business_data_analysis.sql
Install required Python libraries:
pip install -r requirements.txt
Run the Python script to perform analysis and machine learning:
python ml_business_insights.py
- Clear identification of high-revenue sellers
- Quantitative measurement of customer repeat behavior
- Insight into product category performance and trends
- Demonstration of an end-to-end analytics pipeline using SQL and Python
- Business-ready insights that can be used for strategic planning
- Integration of advanced machine learning models
- Automated data pipelines from database to Python
- Dashboard creation using Streamlit or Power BI
- Deployment of analytics as a web service or API
Rahul Patil
Data Analyst and Machine Learning Enthusiast