telco-customer_analysis

About

This is a Mini-Project for SC1015 (Introduction to Data Science and Artificial Intelligence) which focuses on Customer Tenure rate from Telecom Customer Dataset. For detailed walkthrough, please view the main source code in this order:

1. Exploratory Analysis 1
2. Exploratory Analysis 2
3. Data Cleaning & Data Preparation
4. Machine Learning

Contributors

@ir2fan , @ahmadazfar , @MachoPanko

Problem Definition

• Can we find the variables that have the strongest relationship with Customer Tenure?
• Can we produce a model that can accurately predict customer tenure based on certain variables?

Models Used

1. Linear Regression
2. Decision trees
3. XG Boosting
4. Gradient Boosting
5. Random Forest

Conclusion

• The top 6 variables of differing natures are found.
• A model with a prediction accuracy of 99.7% can be produced with random forest.
• We learned that linear regression and decision tree regression did not perform well in predicting tenure.
• We learnt three new machine learning models (random forest, XGBoost and gradient boosting) which offers way more accurate results.

References

1. https://www.kaggle.com/datasets/blastchar/telco-customer-churn https://www.grandviewresearch.com/industry-analysis/global-telecom-services-market

2. https://www.heavy.ai/blog/strategies-for-reducing-churn-rate-in-the-telecom-industry#:~:text=The%20average%20churn%20rate%20in,fostering%20customer%20loyalty%20is%20key.

3. https://www.grandviewresearch.com/industry-analysis/global-telecom-services-market