Student Performance Indicator

This is a Machine Learning End-End Project

Project Overview

This project develops a predictive model to assess student performance using machine learning regression techniques. Aimed at educational institutions, the model predicts academic outcomes based on students' demographic and academic data. The repository includes a full suite of scripts for data preprocessing, model training, prediction, and evaluation. Additionally, the project is configured for AWS deployment with CI/CD integration, ensuring scalability and ease of updates. This approach enhances decision-making and strategic planning in education by providing actionable insights into student performance.

Independent Features

Below are the Independent Features used in the Model:

1. gender: Student's gender (male or female)
2. race_ethnicity: Group classification (e.g., group D)
3. parental_level_of_education: Highest education level of parents
4. lunch: Type of lunch received (standard or free/reduced)
5. test_preparation_course: Completion status of a preparatory course
6. physics_score: Score in physics subject
7. chemistry_score: Score in chemistry subject

Algorithms Used:

Model is trained using the below Machine Learning Algorithms and Algorithm giving the highest accuracy is chosen:

1. Linear Regression: Predicts outcomes using linear relationships.
2. Decision Tree: Splits data based on value conditions.
3. Random Forest: Ensemble of decision trees, reduces overfitting.
4. Gradient Boosting: Improves weak models with error-driven updates.
5. AdaBoost: Boosts weak learners into strong ones iteratively.
6. CatBoost: Gradient boosting with categorical data optimization.
7. XGBoost: Optimized gradient boosting with scalability and speed.

Steps Involved

1. Comprehensive data preprocessing and feature engineering
2. Multiple regression algorithms including Linear Regression, Decision Tree, Random Forest, and Boosting Algorithms.
3. Deployment setup for AWS using CI/CD pipelines.

Components

1. Data Ingestion is apart of components module which means reading dataset from databases/file locations

2. Ingested data is transformed inside data_transformation.py

3. Model Training will happen in model_trainer.py

4. Model prediction is occuring in predict_pipeline.py

5. Logger has all the log files

6. Exception handling is taking care by exception.py, exc_info() tells the file name and line number where the exception is occuring

AWS Deployment

1. Utilization of AWS services like IAM, ECR, and EC2.
2. Docker setup on AWS EC2 for container management.
3. GitHub Actions for CI/CD.

Steps for AWS Deployment

1. Create a user in AWS IAM .
2. Create a repository in AWS Elastic Container Registry.
3. Create an instance in AWS EC2.
4. Setup Docker in AWS EC2 instance.
5. Create Runner in GitHub, execute the commands step by step to for Download, Configure, Use the self-Hosted Runner.

Commands for setting up Docker in EC2 :

sudo apt-get update -y

sudo apt-get upgrade

curl -fsSL https://get.docker.com -o get-docker.sh

sudo sh get-docker.sh

sudo usermod -aG docker ubuntu

newgrp docker

Configure EC2 as self-hosted runner:

Run the commands shown after the runner is created in GitHub for Downloading, Configuring and Running the self-hosted runner

Setup github secrets:

AWS_ACCESS_KEY_ID=

AWS_SECRET_ACCESS_KEY=

AWS_REGION = us-east-1

AWS_ECR_LOGIN_URI = demo>> 566373416292.dkr.ecr.ap-south-1.amazonaws.com

ECR_REPOSITORY_NAME = simple-app

Contributing

Contributions to this project are welcome. Please fork the repository and submit a pull request.