Project 5: Prediction of Customer Churn of a Telecommunications Company
-
Sourced and prepared data for analysis with data preparation done to ensure validity, accuracy, completeness, consistency, and uniformity.
-
Performed feature selection and engineering i.e., label and one-hot encoding, SMOTE, normalization, and train-test splitting with a 0.25 test size and random state of 1.
-
Employed modelling techniques such as ADA Boost, Random Forest, SVM, Lasso Regression, Naive Bayes, and Logistic Regression, with the Random Forest model performing best with a score of 92%.
-
Evaluated the performance of each model using precision, recall, F1 score, and accuracy, with the model accurately predicting that one will discontinue utilizing the company’s services 67% of the time and balancing the situation at hand, accurately projecting the outcomes 82% of the time.
-
Evaluated the effectiveness ML classification models using the confusion matrix, with the model predicting that 305 out of 456 customers who will churn were correct.
