Overview / Usage

Predict behavior to retain customers. You can analyze all relevant customer data and develop focused customer retention programs.

The project studies the behaviour of different customers in a telecommunication company. It uses these different customer behaviour to understand what makes customers churn to help retain customers.

Methodology / Approach

The data set was gotten from kaggle open data sets https://www.kaggle.com/blastchar/telco-customer-churn . The data set was analysed using data visualization tools and other data analysis technologies, it was discovered to be an imbalanced data set. Rows with missing data were deleted from the data set. The categorical features of the data set where separated into binary columns and multi columns the binary columns were label encoded while the multi columns were one hot encoded. The features were reduced by using correlation of features to remove one of two features which had an equal to or greater than 0.90 correlation. The reduced features were used to build model which predicted churn and non churn customers. Pipeline was used to select a model with highest accuracy. Logistic regression was used for the model and the metric used were imbalanced data set metrics such as precision, recall, ROC curve. The model had an accuracy of 80%.

Dataset: IBM sample datasets https://www.kaggle.com/blastchar/telco-customer-churn

Technologies Used

programming language: Python

data visualisation tools: plotly, Matplotlib, Seaborn, scikitlearn

data manipulation tools: pandas

​

​

Repository

https://github.com/Awwal01/Churn-Customer-Prediction.git