🧠 Visa Approval Prediction using Machine Learning

📊 Predicting visa application outcomes using multiple ML algorithms

---

🧾 Overview

This project applies supervised machine learning techniques to predict visa approval outcomes using the Easy Visa dataset.
The process involves data preprocessing, feature engineering, model development, and comparative evaluation of various classification algorithms — including Decision Tree, Random Forest, Gradient Boosting, AdaBoost, and Bagging.

---

🎯 Objectives

• 🔍 Understand and explore the Easy Visa dataset
• 🧹 Perform data cleaning, encoding, and feature selection
• 🤖 Train and evaluate multiple ML classification models
• 📈 Compare results using Accuracy, Precision, Recall, and F1-Score
• 🏆 Identify the most reliable and interpretable model
• 💡 Deliver a transparent, reproducible ML workflow

---

🧩 Dataset

• Source: EasyVisa (synthetic dataset for educational use)
• Features: Education, Job Experience, Wages, Employment Region, Job Training Requirements, etc.
• Target: case_status — Visa Approved (1) or Denied (0)

---

⚙️ Project Workflow

Visa_Approval_Prediction/
├── Data/
│ └── easyvisa.csv
├── Source_Code/
│ └── Visa_approval.ipynb
├── Results/
│ ├── confusion_matrix.png
│ ├── classification_report.txt
│ └── model_summary.txt
├── Documentation/
│ └── Project_Report.pdf
├── requirements.txt
├── .gitignore
└── README.md

---

🤖 Machine Learning Models

This project compared multiple ensemble learning algorithms — both Boosting and Bagging — to identify the most effective method for predicting visa approval outcomes.

Boosting techniques (AdaBoost and Gradient Boosting) showed stronger generalization and recall compared to Bagging-based models.

Model	Accuracy	Precision	Recall	F1-Score
AdaBoost	0.7206	0.7495	0.8737	0.8068
Gradient Boosting	0.7369	0.7669	0.8707	0.8155
Bagging Classifier	0.7121	0.7633	0.8249	0.7929
Random Forest	0.7190	0.7676	0.8308	0.7980

✅ Best Model: Gradient Boosting
💡 Demonstrated the highest F1-Score (0.8155) and recall, indicating superior overall balance and predictive performance.

---

⚙️ Setup and Execution

Step 1: Create a Virtual Environment

python -m venv venv
venv\Scripts\activate

Step 2: Install Dependencies

 pip install -r requirements.txt

Step 3: Launch Jupyter Notebook

jupyter notebook

Step 4: Run the Notebook

Open `Source_Code/Visa_approval.ipynb` and execute all cells in order.

📊 Results and Insights

The project evaluated multiple ensemble machine learning models under two major categories Boosting and Bagging to predict visa approval outcomes using applicant and job-related data.

Model	Accuracy	Precision	Recall	F1-Score
AdaBoost	0.7206	0.7495	0.8737	0.8068
Gradient Boosting	0.7369	0.7669	0.8707	0.8155
Bagging Classifier	0.7121	0.7633	0.8249	0.7929
Random Forest	0.7190	0.7676	0.8308	0.7980

✅ Best Model: Gradient Boosting
💡 Achieved the highest F1-Score (0.8155) and strong recall (0.8707), indicating excellent balance between correctly identifying approved visa cases and maintaining high precision.

---

🔍 Key Insights

• Boosting models (AdaBoost, Gradient Boosting) outperformed Bagging methods (Bagging Classifier, Random Forest) across most metrics.
• Gradient Boosting demonstrated superior performance due to its sequential learning approach, which reduces bias and improves generalization.
• High recall values show that the model effectively minimizes false negatives — an important factor when identifying eligible visa approvals.
• Random Forest and Bagging Classifier also achieved competitive performance, confirming that ensemble techniques outperform simple decision trees.

---

💡 Business Interpretation

The results suggest that visa approvals are most strongly influenced by factors such as:

• Education level and work experience of the applicant
• Employment region and company size
• Prevailing wage and full-time job status

The Gradient Boosting model serves as an ideal decision-support tool for organizations like EasyVisa to predict visa outcomes more accurately, fairly, and efficiently.

---

👩‍💻 Author

Name: Hasnath Unnisa
Email: unnisahasnath@gmail.com
LinkedIn: www.linkedin.com/in/hasnath22