Predicting Customer Churn Using Logistic Regression with Python

Customer churn is a major problem for businesses, as it leads to a loss of revenue and customers. Customer churn (or customer attrition) refers to the loss of customers or subscribers for any reason at all. Businesses measure and track churn as a percentage of lost customers compared to the total number of customers over a given time period. This metric is usually tracked monthly and reported at the end of the month. Therefore, it is important to identify the customers who are likely to churn and take appropriate action to retain them. In this article, we will use logistic regression to predict customer churn using a telecommunications dataset.

Test Dataset: We will use the Telco Customer Churn dataset from Kaggle, which contains information about customers who churned and those who did not. The dataset contains 7043 rows and 33 columns.

Exploratory Data Analysis: We will first explore the dataset to get a better understanding of the data. We will use pandas and seaborn for visualization.

The output shows that the dataset contains no missing values and that there are more customers who did not churn than those who did.

Feature Engineering: We will create new features from the existing features to improve the performance of the model. We will create a new feature, “Total Charges”, by multiplying the “Tenure Months” feature with the “Monthly Charges” feature. The reason for creating this new feature is that the original “Tenure Months” feature was continuous and had a wide range of values, which could make it difficult for the model to learn patterns from it. By creating “Total Charges”, we converted the continuous feature into a categorical one, which could help the model to learn better.

Additionally, “Total Charges” could capture non-linear relationships between the “Total Charges” feature and the target variable “Tenure Months”. For example, customers who have been with the company for a short period of time may be more likely to churn than those who have been with the company for a longer period of time. This information can be useful to the model in predicting customer churn. This is just one example of using an engineered feature for the model.

Logistic Regression: We will use logistic regression to predict customer churn. We will first split the dataset into training and testing sets and then fit the logistic regression model on the training set.

Confusion Matrix:

A confusion matrix is a table that summarizes the performance of a classification model by comparing the actual target values to the predicted values. It is typically represented as a 2x2 matrix for binary classification problems, where the rows represent the actual class labels and the columns represent the predicted class labels. Each cell in the matrix represents the number of instances that belong to a particular combination of actual and predicted classes. The four possible outcomes are:

The confusion matrix can be used to calculate various performance metrics, including accuracy, precision, recall, and F1 score.

Classification Report:

A classification report is another tool for evaluating the performance of a classification model. It provides a summary of several important performance metrics, including precision, recall, F1 score, and support, for each class in the target variable.

In general, a good classifier should have high precision, recall, F1 score, and large support (i.e., the model should be able to predict all classes effectively).

Our Results Summary: The output shows that the logistic regression model achieved an accuracy of 75%. The confusion matrix and classification report provide information about the true positive rate, false positive rate, true negative rate, and false negative rate.’ For class 0, the precision is 0.77, which means that out of all the instances predicted as class 0, 77% were actually class 0. The recall is 0.95, which means that out of all the actual instances of class 0, 95% were correctly predicted by the model. The F1 score is 0.85, which is the harmonic mean of precision and recall. It indicates that the model is reasonably good at predicting class 0, with a good balance between precision and recall.

For class 1, the precision is 0.62, which means that out of all the instances predicted as class 1, 62% were actually class 1. The recall is 0.21, which means that out of all the actual instances of class 1, 21% were correctly predicted by the model. The F1 score is 0.31, indicating that the model is not able to predict class 1 effectively.

In conclusion, the model is good at predicting class 0, but not class 1. This could be due to class imbalance or other issues with the data or model. It might be necessary to perform further analysis and improvements to the model in order to better predict Class 1. The accuracy of the model is 0.75, which means that the model correctly predicted 75% of the instances in the testing set. This is a bit misleading considering that the objective of the model is to correctly predict the number of customers that churn, which the model only achieved a 21% accuracy on.

In summary: In this article, we used logistic regression to predict customer churn using the Telco Customer Churn dataset. We first explored the dataset and then created a new feature. We then fit the logistic regression model on the training set and evaluated its performance on the testing set. The logistic regression model achieved an accuracy of 75% but was only able to estimate 21% of customer churn. Further work can be done to improve the performance of the model by using other machine-learning algorithms and fine-tuning the hyperparameters.

Related posts:

Chakra healing is a spiritual and holistic practice that involves the use of various techniques to balance and align the body’s energy centers or chakras. Chakras are believed to be spinning wheels…

Earlier during the year, I attended Washington State Fair yet another time but first time in the spring. However we’d been longing to attend Evergreen State Fair where the last time we went was in…