Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/jamesehiabhi/connecttel-churn-prediction

Customer Churn Prediction: Unlocking Retention Strategies with Data Science
https://github.com/jamesehiabhi/connecttel-churn-prediction

churn-prediction confusion-matrix exploratory-data-analysis jupyter-notebook logistic-regression machine-learning pdf power powerpoint python

Last synced: 2 days ago
JSON representation

Customer Churn Prediction: Unlocking Retention Strategies with Data Science

Awesome Lists containing this project

README 

        
# ConnectTel-Churn-Prediction

## Customer Churn Prediction: Unlocking Retention Strategies with Data Science



Displays



### Introduction

In the highly competitive telecommunications industry, customer retention is crucial for sustaining revenue growth. Customer churn, the loss of subscribers to competitors is a pressing challenge. This report presents a data-driven approach to understanding and predicting customer churn using machine learning techniques. By leveraging real-world customer data, we uncover key insights that can help **ConnectTel Telecom Company** improve customer retention strategies.



### Table of Contents

- [Problem Definition](#problem-definition)

- [Understanding the Dataset](#understanding-the-dataset)

  - [Key Features in the Dataset](#key-features-in-the-dataset)

  - [Data Cleaning and Preparation](#data-cleaning-and-preparation)

- [Key Insights from Exploratory Data Analysis (EDA)](#key-insights-from-exploratory-data-analysis-eda)

  - [Correlation Analysis](#correlation-analysis)

- [Predictive Modeling](#predictive-modeling)

  - [Feature Engineering](#feature-engineering)

  - [Model Selection](#model-selection)

  - [Handling Class Imbalance](#handling-class-imbalance)

  - [Model Training and Validation](#model-training-and-validation)

  - [Model Evaluation](#model-evaluation)

- [Actionable Insights](#actionable-insights)

- [Business Implications & Recommendations](#business-implications-and-recommendations)

- [Conclusion](#conclusion)



### Problem Definition

- ConnectTel needed a solution to predict customer churn effectively to maintain its customer base.

- Develop a machine learning model that accurately identifies customers likely to churn, focusing on reducing false negatives.



### Understanding the Dataset

The dataset analyzed contains **7,043 customer records** with **20 key features**, including demographics, service subscriptions, billing information, and customer tenure. The primary objective is to predict whether a customer will churn based on these attributes.



#### Key Features in the Dataset:

- **Demographics:** Gender, Senior Citizen, Partner, Dependents.

- **Service-Related:** Phone Service, Internet Service (DSL/Fiber), Streaming Services, Tech Support.

- **Contractual:** Contract Type (Month-to-Month, One-Year, Two-Year), Payment Method, Paperless Billing.

- **Financial:** Monthly Charges, Total Charges.

- **Target Variable:** "Churn" (Yes/No) — indicating whether a customer has left the service.



#### Data Cleaning and Preparation

- I handled missing values using median value of the affected feature to replace the missing values, and ensured that features were well-prepared for analysis.

- The dataset was clean and ready for in-depth analysis and modelling.



Displays



### Key Insights from Exploratory Data Analysis (EDA)

  1. **Churn Rate Trends:**

     - Customers on **month-to-month contracts** show a significantly higher churn rate.

     - Long-term contracts (one-year, two-year) have lower churn rates, suggesting that contract stability plays a key role in retention.



  2. 	**Impact of Monthly Charges:**



     	- Customers with **higher monthly charges** tend to churn more often.

     	- The average monthly charge among churned customers is noticeably higher than retained customers, suggesting pricing sensitivity.



  3. **Senior Citizens & Churn:**

   

      - **16.2% of customers are senior citizens,** and they have a slightly higher churn rate.

      - This suggests that additional engagement strategies (such as tailored promotions or better customer support) might help retain senior customers.



  4. **Service & Feature Influence:**



      - **Tech Support & Online Security are critical factors in retention**.

      - Customers without these add-ons churn at a much higher rate, emphasizing the need to encourage customers to adopt these services.



Displays

Displays



#### Correlation Analysis

A correlation matrix revealed that tenure and TotalCharges have a strong positive correlation, indicating that long-term customers tend to spend more. Conversely, MonthlyCharges showed a moderate positive correlation with churn, suggesting that higher monthly charges might be a factor in customer attrition.



Displays



### Predictive Modeling

#### Feature Engineering

- **Categorical Encoding:** Employed LabelEncoder to encode categorical variables, ensuring the numerical nature of features was preserved.

- **Feature Creation:** Developed new features to enhance the dataset.

- **Effectiveness:** The engineered features significantly improved the models' ability to learn from the data.



#### Model Selection

Eight (8) machine learning models were evaluated, including:

- **Logistic Regression Model**

- **Support Vector Classifier (SVC) Model**

- **XGBClassifier Model**

- **Naive Baye Classifier**

- **SGDClassifier Model**

- **KNeighborsClassifier Model**

- **DecisionTreeClassifier Model**

- **SVC Mode**



**Performance Metrics:**

- **Accuracy:** Measures overall correct predictions.

- **Recall:** Identifies how well we capture actual churned customers.

- **Precision:** Ensures that predicted churns are true positives.

- **F1-Score:** A balance of precision and recall.

The **best-performing model** provided a strong balance between precision and recall, making it suitable for business applications.



#### Handling Class Imbalance

To address the class imbalance, techniques such as SMOTE (Synthetic Minority Over-sampling Technique) and class weighting were employed.



#### Model Training and Validation

- **Initial Model Training:** Established a baseline for model performance.

- **Feature Scaling:** Applied standard scaling techniques to enhance model performance.

- **Class Imbalance:** Addressed class imbalance using techniques such as class weighting, with a focus on improving recall for the positive class (churn).

- **Hyperparameter Tuning:** Conducted hyperparameter tuning to achieve an optimal balance between precision and recall.



#### Model Evaluation

The models were evaluated using confusion matrices, precision, recall, and F1 score, with a focus on reducing false negatives. The Logistic Regression model stood out for its effectiveness in identifying churn cases, aligning well with ConnectTel’s objectives.

- **Accuracy:** 79.7%

- **Precision:** 85.0%

- **Recall:** 88.5%

- **F1-score:** 86.7%

- **AUC-ROC:** 70.9%



Displays



### Actionable Insights

1.	**Retention Strategies for New Customers:** Given that customers with shorter tenures are more likely to churn, targeted retention strategies such as personalized offers and proactive customer support should be implemented for new customers.

2.	**Review Pricing Plans:** The positive correlation between MonthlyCharges and churn suggests that customers find higher monthly charges less appealing. Offering more flexible and competitive pricing plans could help reduce churn.

3.	**Encourage Long-term Contracts:** Customers with longer-term contracts are less likely to churn. Incentivizing customers to opt for annual or biennial contracts could enhance customer retention.

4.	**Enhance Service Offerings:** Features like OnlineSecurity, TechSupport, and StreamingTV showed varying impacts on churn. Enhancing these services and ensuring customer awareness could improve satisfaction and reduce churn.



### Business Implications and Recommendations



- **Encourage Long-Term Contracts:** Offer incentives like discounts or exclusive services to motivate customers to switch from month-to-month to annual contracts, enhancing customer retention.

- **Enhance Customer Support & Tech Services:** Provide free trials or bundle tech support and online security services into plans to reduce churn and improve customer satisfaction.

- **Implement Targeted Retention Campaigns:** Use predictive modeling to identify at-risk customers early and deploy personalized outreach programs (emails, calls, promotions) to boost loyalty and retention.



________________________________________

### Conclusion

Predicting customer churn is essential for telecom businesses to improve retention strategies. By leveraging data science and predictive modeling, **ConnectTel Telecom Company** can proactively identify at-risk customers and implement targeted strategies to enhance customer experience, reduce churn, and boost profitability. This project successfully developed a robust churn prediction model, providing valuable insights and laying a strong foundation for future customer retention efforts. The journey from raw data to actionable insights highlights the importance of data-driven decisions and the power of machine learning in real-world applications.🚀🚀🚀

________________________________________







### *Kindly share your feedback and I am happy to Connect 🌟*



Displays