Bank Term Deposit Prediction

This project aims to predict bank term deposits using various machine learning algorithms.

Prerequisites

The following R libraries are required:

• class
• e1071
• caret
• rpart.plot
• ggplot2
• ranger
• dplyr
• corrplot
• pROC
• reshape2
• shiny
• xgboost

Dataset Description

• File: bank-full.csv

• Description: Main dataset used for training and evaluation.

• File: DatasetTable.csv

• Description: Provides a detailed description of the attributes present in the main dataset.

Data Pre-processing

1. Check for missing values in the dataset.
2. Identify and handle duplicated rows.
3. Convert the target variable 'y' to a binary format (0 for "no" and 1 for "yes").

Exploratory Data Analysis (EDA)

1. Plot histograms for numeric attributes.
2. Visualize the distribution of the target variable 'y'.
3. Generate a correlation matrix for numeric variables.
4. Display bar plots for categorical variables.

Model Building and Evaluation

The following models are trained and evaluated:

Logistic Regression

• A generalized linear model (GLM) with a binomial family.

Decision Tree

• A recursive partitioning method using the rpart library.

Random Forest

• An ensemble learning method that constructs a multitude of decision trees at training time.

The results of the models are then compared in terms of accuracy, sensitivity, specificity, and balanced accuracy.

Visualization

1. Feature importance from the Random Forest model.
2. Comparison of model performance metrics using bar plots.

Acknowledgement

The Portuguese bank dataset used for this project came from the following paper:

• Moro, S., Laureano, R. and Cortez, P. (2011). Using data mining for bank direct marketing: An application of the crisp-dm methodology.