Telecom-Churn-Analysis

OVERVIEW: This project aims to analyze customer churn in the telecom industry, identify key factors contributing to churn, and develop a machine learning model to predict future churn. Additionally, a Power BI dashboard provides visual insights into the data.The workflow demonstrates an end-to-end data analysis pipeline, starting from data extraction and transformation to visualization and machine learning.

TOOLS AND TECHNOLOGIES-

• SQL: Querying and data exploration.
• MS Excel : Additional data manipulation and summarization
• Python: Data cleaning, analysis, and machine learning
• Google Colab: Building and training the machine learning model
• Power BI: Interactive data visualization.

PROJECT WORKFLOW- This project involves multiple stages, from data cleaning to machine learning and visualization. Below is the complete workflow:

1. Data Cleaning and ETL (SQL Server)

• The original dataset was cleaned and processed using SQL Server.
• Performed the ETL (Extract, Transform, Load) process to prepare the data for analysis.
• Data cleaning included handling null values, removing duplicates, and ensuring consistency.

2. Data Manipulation and Views (Excel)

• After processing in SQL, two views were created in Excel for further analysis.
• These views summarized key insights and prepared the data for machine learning.

3. Machine Learning Model (Google Colab)

• Built a Random Forest Classifier using Python in Google Colab to predict future churners.
• Key steps: * Performed data encoding to prepare categorical variables for the model. * Evaluated the model using metrics like confusion matrix and classification report. * The model predicted churn probabilities and saved the predictions to a CSV file for further use.

4. Data Visualization (Power BI)

• Created two dashboards using Power BI:

   Dashboard 1: Insights and trends based on the original dataset.
  

   Dashboard 2: Visualization of the churn prediction results from the ML model.

• Enhanced visuals with filters, slicers, and drill-down options to make the insights actionable.

HOW TO RUN THIS PROJECT-

1. SQL SERVER-

• Import the raw dataset into SQL Server.
• Run the ETL and data cleaning scripts provided in

2. MACHINE LEARNING MODEL-

• Open the python script in Jupyter Notebook or open the Colab file directly.
• Install required libraries.
• Run the notebook to train the model and generate predictions

3. POWER BI-

• Open the file in Power BI desktop.
• Ensure SQL Server is connected to Power BI for live data refresh. *Review interactive dashboards for insights.

KEY INSIGHTS-

• Overall Churn Rate is 27%.
• Female customers contribute 35.85% of the churn rate whereas male customers contribute 64.15% of the churn rate.
• Customers aged 50+ have the highest churn rate at 31% whereas younger customers (under 20) have the lowest churn rate at 23%.
• Top states with the highest churn rates include: Jammu & Kashmir: 57% Assam: 38% Jharkhand: 35%
• Month-to-Month contracts lead to the highest churn (47%) and two-Year contracts have the lowest churn rate (3%).
• Predicted Churn Overview: Total Predicted Churners: 378 Predicted Female Churners: 246 Predicted Male Churners: 132
• Customers with Month-to-Month contracts dominate the predicted churners (355 churners).
• Credit Card payment users have the highest predicted churn (192), followed by "Bank Withdrawal" (150).
• Customers with less than 6 months of tenure exhibit the highest predicted churn (65 churners) whereas tenure of more than 24 months shows 106 predicted churners.
• Uttar Pradesh, Maharashtra, and Tamil Nadu have the highest number of predicted churners (44, 40, and 37, respectively).