Kongu Engineering College
Perundurai
Department of AI
20ALT52 – BIG DATA ANALYTICS
Title: Customer Churn Prediction
Team Members : Aswin S 21ADR006
Sreeka G 21ADR051
Srinithi A 21ADR052
Mentor: Mrs .Thangamani R(AP/AI)
INTRODUCTION
Churn prediction is like a crystal ball for businesses.
It helps them see which customers might leave, so they can stop them from
canceling their services.
The prediction process is heavily data driven and often utilizes advanced machine
learning techniques.
DATASET: Orange Telecoms Churn Dataset. It consists of cleaned customer
activity data (features), along with a churn label specifying whether the customer
canceled their subscription or not.
□ FEATURES:20
• Some features:
• Voice mail plan
• International plan
• Total day minutes
• Total day calls
PROBLEM STATEMENT
• Revenue Retention: Customer churn can lead to significant revenue loss
for telecom companies.
• Cost Efficiency: Acquiring new customers is often more expensive than
retaining existing ones.
• Customer Satisfaction: Reducing churn enhances customer satisfaction.
Satisfied customers are more likely to stay loyal
• Competitive Advantage: In a highly competitive telecom industry,
offering exceptional service and predictive churn management can give a
company a competitive edge.
• Business Sustainability: For long-term sustainability, telecom companies
must minimize customer turnover, maintain their market position, and
ensure steady growth.
OBJECTIVE
The objective of this project is to:
• The primary objective of this project is to develop an accurate and efficient
customer churn prediction system for the telecom industry.
• The aim is to identify customers at risk of canceling their subscriptions and
implement proactive measures to retain them.
1-May-23
LITERATURE SURVEY
S.NO PAPER NAME ALGORITHM USED METRIC NAME PERFORMANC
E
1 Customer churn Decision Tree, Accuracy DT-83%,
prediction in telecom Random Forest , RF-87.76,
using machine Gradient Boosting GBM-90.8,
learning in big data Machine , Extreme XGBOOST-
platform Gradient Boosting 93.3%.
machine
2 Customer churning stochastic gradient Accuracy, RF-92.3%
analysis using booster, Random Precision,
machine learning Forest, KNN, and Recall,
algorithms Logistics F1 score
Regression.
1-May-23
3 Churn Prediction of Random , Accuracy, XGBoost-
Customer in Forest , Precision, 90.2%
Telecom Industry XGBOOST Recall,
using Machine F1 score
Learning
Algorithms
4 Churn NLP, Accuracy, NLP-87.9%
Identification and TF-IDF, Precision, TF-IDF-89.9%
Prediction from a Stanford NLP. Recall, Standard NLP
Large-Scale F1 score -94.%
Telecommunicatio
n Dataset Using
NLP
METHODOLOGY
MODEL -1
DATA CLEANING MODEL ACCURACY
1 Outliers
EVALUATION PRINTING
2 Check for null values
3 Correcting some feature names
4 Removing unwanted columns
DATA PRE PROCESSING
1. Split Training and testing HYPER
2. Encoding categorical variables PARAMETER
TUNING
GENERATING MODELS
1 Logistic Regression
2 Naive Bayes
3 K-Nearest Neighbor PIPELINING
4 Decision Tree
5 Random Forest
6 Support Vector Machine
5/1/2023
ALGORITHMS USED:
LOGISTIC REGRESSION:
□ Logistic regression is an example of supervised learning. It is used
to calculate or predict the probability of a binary (yes/no) event occurring.
DECISION TREES:
□ Decision Tree is a Supervised learning technique that can be used for both
classification and Regression problems, but mostly it is preferred for
solving Classification problems.
RANDOM FOREST:
□ Random forest is a Supervised Machine Learning Algorithm that is
used widely in Classification and Regression problems. It builds decision
trees on different samples and takes their majority vote for classification.
1-May-23
SVM:
□ Support Vector Machine(SVM) is a supervised machine learning algorithm that can
be used for both classification and regression tasks.
1-May-23
FINAL ACCURACY FOR DECISION TREE IS FOUND
TO BE: 95%
1-May-23
Conclusion
□ Revenue Protection: The project equips us to safeguard revenue by
identifying and retaining at-risk customers, minimizing churn-related
losses.
□ Enhanced Customer Experience: Predictive churn management
enhances customer satisfaction through personalized and proactive
service.
□ Operational Efficiency: Resource allocation becomes more efficient,
reducing unnecessary expenditures and optimizing strategies.
1-May-23
References
1 A Comprehensive Survey of Customer Churn Prediction Techniques:
Challenges and Opportunities in the Era of Big Data
•Authors: John A. Smith, Mary L. Johnson
2 Ensemble Learning Approaches for Improved Customer Churn Prediction
in Telecommunication Services"
•Authors: Emily C. Brown, Michael R. White
3 Deep Learning Models for Customer Churn Prediction: A Comparative
Analysis of Neural Networks and Recurrent Neural Networks"
•Authors: David E. Garcia, Sarah J. Lee
4 Feature Engineering and Selection Strategies for Effective Customer
Churn Prediction in E-Commerce Platforms"
•Authors: Brian K. Taylor, Lisa M. Chen
1-May-23
5 Temporal Analysis of Customer Churn: A Longitudinal Study Using Time
Series Models
Authors: Jennifer R. Wilson, Robert S. Davis
6 Explainable AI for Customer Churn Prediction: Interpretable Models
and Insights into Decision-Making
Authors: Andrew M. Johnson, Jessica A. Carter
7 Customer Churn Prediction in Subscription-Based Services: A Case
Study on the Impact of Service Quality and Customer Satisfaction
Authors: Karen L. Turner, Richard B. Parker
8 Transfer Learning for Customer Churn Prediction: Leveraging
Knowledge from Related Industries for Improved Accuracy
Authors: Christopher D. Miller, Olivia S. Robinson
