Utilizing Logistic Regression for Telco Customer Churn Prediction
After conducting an exploratory analysis, use the previous dataset to make a prediction model using Logistic Regression. Surely, had to check back on a dataset to data in the form of text converted into numerical data for models to be extracted.
Logistic Regression is a classification algorithm to find the relationship between features (input) discrete/continuous with the probability of certain discrete output results. Binary Logistic Regression: This is a Logistic Regression that only has 2 outputs (classified into two different classes). This experiment will try to predict the two outputs (Yes-No).
Several things need to be prepared in making this model, of course, use python and enable Pandas and Numpy libraries for data analysis tools, Matplotlib and Seaborn libraries for data visualization, Scikit-Learn Library for Machine Learning.
Using multiple independent variables as the training dataset for machine learning, Tenure Months, Internet Services, Online Security, Online Backup, Device Protection, Tech Support, Monthly Charges, Total Charges.
Creating Logistic Regression modeling by using these variables and testing to see if the model is feasible for use in the prediction. Using measurement to check with precision, recall, and accuracy. Obtained an accuracy of 80% for the prediction model to be used.
Conduct experiments to test the predictions of the model with the following conditions. The results will determine whether the customer with the specified conditions is included as a churned customer.
Then the results obtained prediction has been made that a customer with such conditions as indicated above churn worth of 0, which means that customers do not churn when the prediction results show 1 means that customer churn.
These models can then be used to make predictions but it needs to be reconsidered because of its accuracy by 80%. If the accuracy of this model is above 90%, it will be categorized as better and more accurate.
