Youtube2 Project Overview by Nitesh

Introduction

• Presenter: Nitesh
• Topic: Deploying Machine Learning Projects using Render
• Project Type: Hobby-level and portfolio-building projects

Issues with Previous Platform (Roku)

• Roku previously offered a free-tier service, making it accessible for individuals without a company setup.
• Roku's free tier has been discontinued.
• New deployment on Roku now requires payment, which is problematic for students or hobbyists.

Solution: Render

• Render Platform: Offers a free tier similar to Roku
• Suitable for hobby and portfolio projects
• Easier deployment compared to Roku
• Handles machine learning projects efficiently

Step-by-Step Guide

Step 1: Machine Learning Model Preparation

1. Toy Problem: Predicting student placement based on features (IQ, CGPA, 10th & 12th marks, communication skills).

2. Dataset: Toy dataset with 100 rows.

3. Model Used: Random Forest Classifier

4. Process:

   • Import libraries
   • Export library versions for consistency
   • Import and prepare data
   • Split data into training and testing sets
   • Hyperparameter tuning using Grid Search CV
   • Train and select the best model
   • Save model using pickle

5. Model Results:

   • Accuracy around 40%
   • Emphasis on having a working model rather than high accuracy

Step 2: Creating Deployment Setup Using Flask

1. Creating Virtual Environment:

   • Create project folder and initialize virtual environment (myenv)
   • Activate the virtual environment

2. Project Files:

   • .gitignore: To exclude myenv
   • Model file: Upload model.pkl (pickle file)
   • app.py: Flask application file for handling requests and loading model
   • index.html: Template file for user interface

3. Installing Libraries:

   • Required libraries: Flask, numpy, pandas, scikit-learn, gunicorn
   • Create requirements.txt using pip freeze > requirements.txt

4. Testing Locally:

   • Run python app.py to start Flask server
   • Test application locally by entering student details and predicting placement

Step 3: Uploading to GitHub

1. GitHub Repository Setup:
   • Create a new repository on GitHub
   • Add repository as remote to the project
   • Push local project files to GitHub

Step 4: Deploying on Render

1. Render Account Setup:

   • Sign in to Render using GitHub
   • Connect GitHub repository to Render

2. Deployment Process:

   • Create a new web service on Render
   • Select GitHub repository and configure deployment settings
   • Render will automatically install dependencies and start the server

3. Final Testing:

   • Access the deployed application via Render-provided URL
   • Test by entering student details and checking placement prediction

Summary and Conclusion

• Render offers a free and easy-to-use platform for deploying machine learning projects.
• Suitable for students and hobbyists without budget for paid services.
• Detailed deployment steps including model preparation, Flask setup, GitHub integration, and Render deployment.
• Encouragement to explore and utilize Render for personal projects.