A European Banking Institution has provided data related to direct marketing campaigns. The campaigns were based on phone calls to their customers in order to offer term deposit subscriptions.
In this project, we will build classification models to predict whether the client will subscribe to a term deposit or not and also to identify the main factors that affect the clients’ decision.
The data used for this project are publicly available on UCI Machine Learning repository - Bank Marketing Dataset. Further details about the dataset is provided in the README file in the data folder.
Data Cleaning and Exploratory Data Analysis is done on the dataset. The plots and the inferences can be found in the analysis folder.
Feature Selection is done to select the best variables for prediction.A major problem associated with the given dataset is that it is imbalanced. So, SMOTE - a Synthetic Data Generation sampling method is used to overcome this problem.
Feature Engineering folder contains the respective code and explanation.
Four different algorithms are used in this project to predict the term deposit subscription.
- Linear algorithm: Logistic Regression (GLM)
- Non-Linear algorithm: Support Vector Machine (SVM)
- Bagging algorithm: Random Forest (RF)
- Boosting algorithm: eXtreme Gradient Boosting (XGBoost)
The algorithms were assessed using test accuracy, AUC and F1 Score. Random forest performed the best among the four algorithms with 87.41% test accuracy, 71.08% AUC and 0.4723 F1 Score
R codes for the implementation of all the above algorithms and the prediction are provided in the prediction folder.
The prediction on the test data using the Random Forest algorithm is stored in the result.csv file.
R Codes for each of the above steps are provided separately in the respective folders along with the README file and also a complete R Script for this whole project is provided for convenience.