TelcoCustomerChurn_EDA-ML_Python

Project: Programming for Data Science

Topic: Predictive Churn Analysis for Telecom Company

Keywords: Data Science, Telco Customer Churn, Exploratory Data Analysis (EDA), Machine Learning, Predictive Churn Analysis, Classification, Python

Table of Content

• Project Overview
• Motivation
  • Aim & Objective
• Contents
• Technologies Used
• License
• Credits

Project Overview

• Use predictive analytics to identify which telecom customer will churn or not based on features like Gender, SeniorCitizen, MultipleLines, etc.
• The dataset can be obtained here from Kaggle.
• Python is used to assist this project with Data Mining by extracting important insights using:
  • Basic Statistics
  • Data Visualization
  • Exploratory Data Analysis
  • Supervised Machine Learning Models (Classification)
    • Support-Vector Machine (SVM)
    • Logistic Regression
    • Random Forest Classifier (Bagging Ensemble Model)
    • XGBoost Classifier (Gradient Ensemble Model)

Motivation

• The telecom industry will be influenced by the broad availability of offers and incentives from different service providers.
• Because of the variety of types of churn and the different causes for individual turnover, internal teams would fail to comprehend the scarcity of consumer attrition.
• As a result, research has shown that getting new consumers is costly, but losing current customers is much more costly since existing paying customers are often repeated customers who, if happy, would buy and use the goods or services again (Ranabhat, 2018).
• Customer turnover data is constantly changing as many organisations build new data plans and packages to obtain a competitive edge, and enterprises must be careful in monitoring customer behaviour to minimise customer churn by undertaking predictive modelling such as churn analysis (Khodabandehlou & Rahman, 2017).

Aim & Objective

• Aim:
  • To improve the process of analysing customer churn in the telecommunications industry so that it can focus on maintaining long-term relationships with loyal customers while also developing an effective prediction model that divides telecom customers into churners and non-churners.
• Objective:
  • To create and select the best Machine Learning model that classifies telecom customers as actual churners or not based on the importance of data variables and models evaluation and assessment (i.e. Accuracy, Recall, AUC, etc.).
  • The insights and variables gained may be utilised to make better choices and adjustments, such as adding or upgrading services to decrease churn and measuring the success of marketing and other customer acquisition methods and techniques.

Contents

(1) TelcoCustomerChurn_Dataset.csv

• Telecom Customer Churn dataset file in CSV format.

(2) TelcoCustomerChurn_EDA-ML_Python Folder

• Contains 3 Python notebooks with implementation codes and explanations for the project.

  (2.1) TelcoCustomerChurn_EDA-ML_Python (3 LRs).ipynb

  • The notebook containing the Python implementation codes (along with explanations) using 3 Logistic Regressions with different splits for the project.

  (2.2) TelcoCustomerChurn_EDA-ML_Python (RF & LR).ipynb

  • The notebook containing the Python implementation codes (along with explanations) using Random Forest and Logistic Regression for the project.

Technologies Used

License

• None (for now)

Credits

• Took inspiration from Kaggle