https://github.com/jasontanx/deep-learning-bank-deposit

A bank deposit prediction (deep learning) project from my MSc Data Science course
https://github.com/jasontanx/deep-learning-bank-deposit

bank-marketing data-science deep-learning

Last synced: 3 months ago
JSON representation

A bank deposit prediction (deep learning) project from my MSc Data Science course

• Host: GitHub
• URL: https://github.com/jasontanx/deep-learning-bank-deposit
• Owner: jasontanx
• Created: 2023-02-05T14:41:58.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2023-02-07T14:51:19.000Z (over 2 years ago)
• Last Synced: 2025-02-01T02:19:54.410Z (5 months ago)
• Topics: bank-marketing, data-science, deep-learning
• Language: Jupyter Notebook
• Homepage:
• Size: 1.21 MB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# deep-learning-bank-deposit: Bank Marketing 🏦

Topic: **Development of An Enhanced Deep Learning Model to Predict Client's Intention to Subscribe to the Bank's Term Deposit**

No | Dataset | Information

--- | --- | --- 

1 | URL | https://www.kaggle.com/datasets/prakharrathi25/banking-dataset-marketing-targets?select=train.csv

2 | Dataset Name | Portuguese Bank Direct Marketing 

3 | File Type | csv file

4 | Observation | 45,211

5 | Features | 17

6 | Data label | “Yes” referred to bank clients successfully subscribing to the term deposit. “No” referred to bank clients that rejected the subscription.

# Introduction

## A brief overview into the project domain

- Marketing functions have always been playing a central role in the financial industry, especially in the banking sector

- Retail banks often used direct marketing as a telemarketing strategy to contact potential customers and sell their products

- Crucial for retail banks to ensure that they are targeting groups with a high chance of success

## What more could be done?

- Data analysis. Understand the consumers needs and preferences!

- Leverage on deep learning techniques to make better predictions

## What is the problem statement of the project?

- Retail banks urgently need a reliable and accurate machine learning model as a competitive advantage to help them predict customer intention to subscribe to term deposits

- Offerings of financial products like providing “term deposits” slightly vary from the other retail banks. In other words, every bank offerings are almost identical)

## Aims & Objectives (What do I aim to achive?) 🌟

**The Aims**

- The overall aim of this project is to enhance retail banks’ marketing effectiveness and reduce marketing costs through the development of a reliable deep learning machine learning model to accurately predict bank clients’ possibilities in subscribing to bank term deposit. 

**The Objectives**

- To identify features that play a major role in affecting the bank clients’ intention to subscribe to the bank term deposit. 

- To develop a reliable deep learning technique and predict bank clients’ intention to subscribe to a financial product - bank term deposit.

- To evaluate the performance of the deep learning models with the evaluation metrics benchmarked by past studies.

# Initial Data Exploration & Exploratory Data Analysis (EDA)

- Finding out the following: 

  - What is the data shape? 

  - Are there any missing values? 

  - How many categorical / numerical variables are there?

  - What is the dependent variable, how's the distribution? 

  - Are there any class imbalance issue? 

  - and many more...!

- EDA --> Univariate Analysis & Bivariate Analysis

# Data Pre-Processing

![git_5_dl_pre-process](https://user-images.githubusercontent.com/116934441/216932602-7027758a-9ba6-4247-b911-c63c1e620f30.png)

- Correlation Analysis

- Label Encoding

- One Hot Encoding

  - [Label vs. One-Hot Encoding Short Explaination on Kaggle](https://www.kaggle.com/getting-started/187540)

- Data Partitioning

- Class Re-Sampling

- Data Normalisation

- Feature Selection

# Modelling

## Models Developed

- 1 Baseline Model

- 4 ANN Model

- 2 RNN Model

- 1 LSTM Model

## Proposed Deep Learning Model Flowchart 

![git_7_ann_flow_dl](https://user-images.githubusercontent.com/116934441/217275836-c7187894-b789-437c-b155-70ea8a8b35bc.png)

## Hyperparameters Involved

- Learning Rate ✅

- Epoch ✅

- Dropout ✅

- Batch Size ✅

# Performance Evaluation

![git_6_perform_eval)dl](https://user-images.githubusercontent.com/116934441/216936051-bd8b7058-385c-460f-baaa-5db28bd8267f.png)

## Critical Analysis

- Among all the models developed, the highest accuracy of 90.29% ✅ was achieved by model 4 

- Class imbalance issue was resolved with the application of the SMOTE technique

  - [What is SMOTE technique?](https://towardsdatascience.com/smote-fdce2f605729)

- Some evaluation metrics carry more weight as compared to others

- Focus of the retail bank should be on correctly predicting the bank clients that would subscribe to the deposits 

- Hence, high sensitivity or TPR will be much more important

  - Banks prefer to correctly predict clients that would most likely purchase their term deposits

  - Banks stand to lose out more in terms of the sales opportunity if highly potential clients are missed out by the model 

  - On the other hand, banks could afford to wrongly identifying not interested clients as highly likely to purchase