I decided to go with Logistic Regression, after checking the confusion matrix. It gave an accuracy of 80.23%, and my other models also gave similar accuracies.
Accuracy : 0.8023696682464455
Sensitivity : 0.9028938906752412
Specificity : 0.5207207207207207
● As we saw in our EDA, having a 2-year contract, DSL internet services,1, 2,3 Year tenure reduces chances of churn. Generally, 1,2 years have to wait till the end of the contract.
● Lastly, total charges, monthly contracts, fibre optic internet services and 4,5,6 Year tenure can lead to higher churn rates. This is interesting because although fibre optic services are faster, customers are likely to churn because of it.
● Gender of a person has nearly no impact on the churn rates.
● Senior citizen makes to be 16% of the total customer but have a higher churning rate: 42% against the non-senior citizen.
● Customers with no partner or no dependents are likely to churn.
● The ones with multiple lines are more likely to churn than the ones with a single line.
● The ones with paperless billing are more likely to churn.
● To keep customers from churning, focus on making them giving the yearly contract. Give customers attractive offers on yearly contracts.
● Try not to offer all the customers, the fibre optic internet services because it has a high churn rate. Mostly because of the high cost associated with this service.
● Focus on customers with no internet service, they have a very low churn rate.
● Longer the contract, lesser the monthly charges leads to low rates of churning.
As tenure increases, the probability of churn decreases.
● As monthly charges increases, the probability of churn increases. Therefore, try to control it.