📊 Machine Learning Projects Collection

8 complete ML projects covering regression, classification, clustering, deep learning, NLP, and transfer learning.
Each project uses real/standard datasets and is fully runnable.

---

📊 Results Summary

#	Project	Model	Dataset	Key Metric
1	House Price Prediction	Ridge Regression	California Housing	R² ≈ 0.60
2	Spam Classifier	Linear SVM	20 Newsgroups	Accuracy ≈ 98%
3	Customer Segmentation	K-Means (K=4)	Synthetic Blobs	Silhouette ≈ 0.55
4	Image Classification	CNN (PyTorch)	CIFAR-10	Accuracy ≈ 78%
5	Sentiment Analysis	Logistic Regression	20 Newsgroups	F1 ≈ 0.93
6	ANN Diabetes Prediction	PyTorch ANN	Pima-like Synthetic	Accuracy ≈ 76%
7	Transfer Learning	ResNet18 Fine-Tuned	ImageNet → Custom	Feature Extraction
8	Word Embeddings	Word2Vec + t-SNE	Custom Vocabulary	Cosine Similarity

---

📁 Project Structure

ml-projects-collection/
├── 01_house_price_prediction/
│   └── main.py          # Linear, Ridge, Lasso comparison
├── 02_spam_classifier/
│   └── main.py          # TF-IDF + Naive Bayes + SVM
├── 03_customer_segmentation/
│   └── main.py          # K-Means + PCA visualization
├── 04_image_classification/
│   └── main.py          # CNN on CIFAR-10 (PyTorch)
├── 05_sentiment_analysis/
│   └── main.py          # TF-IDF + Logistic Regression
├── 06_ann_diabetes_prediction/
│   └── main.py          # PyTorch ANN with custom DataLoader
├── 07_transfer_learning/
│   └── main.py          # ResNet18 fine-tuning vs feature extraction
├── 08_word_embeddings/
│   └── main.py          # Word2Vec, cosine similarity, t-SNE
├── requirements.txt
└── README.md

---

🚀 How to Run

git clone https://github.com/imranalimemon/ml-projects-collection.git
cd ml-projects-collection
pip install -r requirements.txt

# Run any project
python 01_house_price_prediction/main.py
python 02_spam_classifier/main.py
python 03_customer_segmentation/main.py
python 04_image_classification/main.py
python 05_sentiment_analysis/main.py
python 06_ann_diabetes_prediction/main.py
python 07_transfer_learning/main.py
python 08_word_embeddings/main.py

---

💡 Key Learnings

1. Feature scaling is critical for linear models but not for tree-based models
2. TF-IDF + SVM is a surprisingly strong baseline for text classification
3. K-Means requires choosing K carefully — use elbow method and silhouette score
4. CNNs can learn spatial features automatically from images
5. Model comparison is essential — never rely on a single model
6. Custom Datasets & DataLoaders in PyTorch give full control over batching and preprocessing
7. Transfer learning dramatically reduces training time — freeze layers, retrain classifier
8. Word embeddings capture semantic similarity — similar words cluster in vector space

---

📄 License

MIT License

Built by Imran Ali — MUET Jamshoro, CS 2026