Salary Estimation using Artificial Neural Networks (ANN)

A machine learning project that predicts a bank customer's estimated salary based on their demographic and account information. This project features an ANN regression model built with TensorFlow/Keras and deployed as an interactive web application using Streamlit.

🔴 Live Demo: Click here to use the App

📌 Project Overview

The goal of this project is to build a regression model that can estimate a user's salary based on standard banking data points. It demonstrates the end-to-end process of data preprocessing, model training, and deployment.

Key Features:

• Deep Learning Model: An Artificial Neural Network (ANN) regressor trained on the Churn Modelling dataset.
• Web Interface: A user-friendly Streamlit app for real-time predictions.
• Preprocessing Pipeline: Automated handling of categorical data (Label Encoding & One-Hot Encoding) and feature scaling.

🛠️ Tech Stack

• Python (Programming Language)
• TensorFlow & Keras (Model Building)
• Pandas & NumPy (Data Manipulation)
• Scikit-learn (Data Preprocessing)
• Streamlit (Web Framework)
• Pickle (Model Persistence)

📂 Project Structure

Salary-Estimation-Project/
│
├── venv/                          # Virtual environment (do not track in git)
│
├── README.md                      # Project documentation
├── requirements.txt               # List of python libraries needed
│
├── Churn_Modelling.csv            # The dataset
├── salaryregression.ipynb        # Notebook used to train the model
├── app.py                         # The Streamlit web application
│
├── # --- Generated Files (Created after running the notebook) ---
├── regression_model.h5                # The trained ANN model
├── salary_scaler.pkl              # Saved Standard Scaler
├── salary_gender_encoder.pkl      # Saved Label Encoder
└── salary_geo_encoder.pkl         # Saved One-Hot Encoder

🚀 Installation & Setup

1. Clone the repository:

   git clone [https://github.com/yourusername/salary-estimation-ann.git](https://github.com/yourusername/salary-estimation-ann.git)
   cd salary-estimation-ann

2. Create a virtual environment (Optional but recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows use: venv\Scripts\activate

3. Install dependencies:

   pip install streamlit tensorflow pandas numpy scikit-learn

🏃‍♂️ How to Run

1. Train the Model (First time only): Run the salaryregression.ipynb notebook to preprocess the data, train the ANN, and save the necessary .h5 and .pkl files.

2. Start the Streamlit App:

   streamlit run app2.py

3. Access the App: Open your browser and go to http://localhost:8501.

📊 Model Architecture

The ANN consists of:

• Input Layer: Matches the number of preprocessed features.
• Hidden Layers: Two dense layers with ReLU activation (64 and 32 neurons).
• Output Layer: Single neuron with linear activation (for regression output).
• Optimizer: Adam
• Loss Function: Mean Squared Error (MSE)

📝 Usage

Use the sidebar/main panel in the web app to input:

• Geography (France, Germany, Spain)
• Gender
• Age, Tenure, Balance
• Credit Score, Number of Products
• Credit Card Status & Active Membership

Click "Estimate Salary" to see the predicted value.