🧠 Project_Machine_Learning

A long-term ML journey built notebook by notebook — pure consistency, pure curiosity, pure hard work.
Every week I added a little more, learned something new, and slowly stacked up a full collection of machine learning projects that actually do something.

All projects are .ipynb files and can be ported into a Django REST API.
Once wrapped as endpoints → they run 24/7, powering any frontend in any language. All projects are future-proof

---

🎯 Project Overview

Category	Target	Completed
Supervised Learning	10	✅ 10/10
Unsupervised Learning	10	✅ 10/10
Reinforcement Learning	10	🚧 1/10

---

🔵 Supervised Learning (10/10 projects Completed)

1. Linear Regression - House Prices Prediction
2. Logistic Regression - Heart Disease Prediction
3. Decision Trees - Customer Churn Prediction
4. SVM - Handwriting Detection
5. Naive Bayes - Sentiment Analysis
6. K-Nearest Neighbors - Malaria Detection from Blood-cell Images
7. Random Forest - Stock Price Prediction
8. Support Vector Machine - Whiskey Brand Prediction from Flavour
9. Multi-Layer Perceptron - Handwritten Digit Classifier
10. TensorFlow - Denoising Images

---

🔵 Unsupervised Learning (10/10 projects Completed)

1. K-Means Clustering - Customer Segmentation

2. DBSCAN - Anomaly Detection in Network Traffic

3. Document Topic Modeling using LDA

4. Music Genre Discovery (Raw MP3 Clusters)
   This Project was bit sophisticated to fit into a google colab notebook so was done in a folder and uploaded via git. See Project14

   1. . Project 14 - Making an AI model to learn and figuring out music genre from a bunch of raw mp3 files
      1. For example: made the ai model learn to figure out which of which are Enrique songs and AR Rahman songs and Taylor Swift songs and Owlcity songs and other mp3 songs just from a zip containing a bunch of mp3 files.

5. Movie Recommender System - Collaborative Filtering

6. Genetic Clustering of DNA Sequences

7. Dimensionality Reduction for Image Compression Converts a high resolution image to low resolution. And then we train the AI model to analyze both. This process we do for big datasets of upwards of 1000 images. Then once the AI model has learnt and analyzed and understood patterns in the images. We now give the AI model purposefully a low resolution image. And the AI model from what it learnt previously tries to build a high-resolution image out of the purposefully given low resolution image.

8. Fraud Detection in credit card transactions using Autoencoders Built an AI model that analyzes a credit card database that contains upwards of 2.84 lakh transactions. And from that it first learns to analyze the transactions. Then finds out which of those transactions were fraudulent transactions. In the output we received in the final output. 492 transactions were fraud transactions out of the total 2.84 lakh transactions! Honestly it feels kind of brilliant to know what these AI models can be taught to learn and perform.

9. Social Network Graph Analysis using Community Detection In this project we built an AI model that analyzes a social group and identifies patterns in their friend circles. We used a unsupervised learning AI technique called community detection to analyze the different friends gangs formed and made the AI model understand and discover which people belong together. The model visually shows who becomes friends with who.

10. Market Basket Analysis using Association Rule In this project we built an AI model that analyzes the customer data from a supermarket's past customer purchase data and then figures out patterns from it. It finds out things like below

• When customers purchase WHOLE MILK, 1.74% also purchase HAM
• When customers purchase CHOCOLATE, 11.90% also purchase ROLLS/BUNS
• When customers purchase FRUIT/VEGETABLE JUICE, 11.00% also purchase ROLLS/BUNS

and so the AI model understands past customer purchases and figures out patterns in which future customers might also behave. thereby, increasing company revenue.

---

🔵 Reinforcement Learning (1/10 Completed)

1. Self-Learning Tic-Tac-Toe (Q-Learning)
   A simple agent that learns entirely through rewards and exploration — no rules hardcoded.

---

🎧 Additional Projects

Proj 1: “Hello Voice” — English TTS

Text → English mp3 voice.

Proj 2: “Namaskaram Malayalam” — Malayalam TTS

Malayalam text → Malayalam voice.

Proj 3: “Panchayath Podcast” — Malayalam Podcast-Style TTS

Long-form Malayalam narration generation.

Proj 4: Voice VAE — Generate Spectrograms

Variational autoencoder for creating synthetic spectrograms.

**[Proj 5: Machine learning vs Deep Learning]

(https://github.com/ruforavishnu/Project_Machine_Learning/blob/master/rose_vs_carnation_ml_vs_dl.ipynb)**

Machine Learning vs Deep Learning complete blog post : (https://dev.to/vishnu_ajit/4-reasons-why-ditching-machine-learning-and-falling-in-love-with-deep-learning-might-be-a-good-idea-3lm1)

---

🔧 Tech Design

• All projects are deployable via Django REST API
• Written as colab notebooks to save hard-disk memory when doing multiple pip installs for each project
• ML models can be exported as pickle or joblib as and when necessary
• Build the backend in Python and expose its REST API and it shall run in any front-end no matter what technology comes tomorrow is the basic philosophy. Connect the backend REST API to the front end and it shall run 24/7 like clockwork.
• 🚀 Everything is future-proof.

---

💬 Final Note

This repo is not meant to look fancy — it’s meant to show a consistent journey and pure hardwork , built week after week.

Now updated new readme.md with better structure, badges, and hyperlinks, it carries the same weight as the effort behind it.

---