📘 MACHINE LEARNING ASSIGNMENTS – UML501

Mehak | B.Tech COE (TIET)

This repository contains a collection of 8 Machine Learning assignments completed as part of coursework. Each assignment includes hands-on implementation of ML algorithms, reinforcing both theoretical knowledge and practical programming skills using Python.

---

📂 Assignment Overview

Assignment 1 – NumPy Operations & Matrix Computations

Objective: Introduction to numerical computing using NumPy
Key Tasks:

• Array creation, slicing, reshaping, flattening
• Matrix operations (addition, multiplication, inverse, determinant, eigenvalues)
• Statistical measures (mean, median, SD, covariance, percentiles)
• Image-to-array conversion & file handling

---

Assignment 2 – Data Preprocessing Techniques

Objective: Clean and transform raw data for ML models
Key Tasks:

• Handling missing values & noise removal
• Normalization & standardization
• Binning & discretization
• One-hot encoding, ordinal encoding
• Similarity & correlation metrics (Jaccard, Cosine, Pearson, Simple Matching) Dataset: Bike Buyers Dataset (synthetic equivalent)

---

Assignment 3 – Regression Models & PCA

Objective: Compare analytical & iterative regression training
Key Tasks:

• Linear Regression using Normal Equation + Gradient Descent
• 5-Fold Cross Validation evaluation
• Model performance comparison via R² score
• PCA for dimensionality reduction (before vs after comparison)

---

Assignment 4 – Web Scraping & Data Extraction

Objective: Collect real-world structured data
Key Tasks:

• Static scraping using BeautifulSoup
• Dynamic scraping using Selenium
• Extracted data from:
  • BooksToScrape
  • IMDb Top 250 Movies
  • TimeAndDate Global weather reports
• Export to CSV for analysis

---

Assignment 5 – Ridge & Lasso Regression + Cross Validation

Objective: Regularized regression & model selection
Key Tasks:

• Ridge Regression using Gradient Descent (with tuning of α & LR)
• Linear vs Ridge vs Lasso comparison
• RidgeCV & LassoCV on Boston Dataset
• Hitters Dataset regression evaluation and best model justification

---

Assignment 6 – Naïve Bayes & GridSearchCV

Objective: Bayesian modeling and hyperparameter tuning
Key Tasks:

• Gaussian Naive Bayes – manual & in-built implementation
• GridSearchCV for best K in KNN

---

Assignment 7 – Support Vector Machines

Objective: Classification using different SVM kernels
Key Tasks:

• SVC with Linear / Polynomial / RBF kernels
• Metrics: Accuracy, Precision, Recall, F1-score
• Confusion Matrix visualization
• Effect of feature scaling on SVM performance

---

Assignment 8 – AdaBoost (Text, Medical & Sensor Data)

Objective: Boosting for stronger ensemble models
Parts Implemented:

1. SMS Spam Classification
   • TF-IDF vectorization + manual AdaBoost (T=15) + sklearn AdaBoost
2. Heart Disease Prediction
   • UCI Heart dataset with hypertuning of estimators & learning rate
3. WISDM Smartphone & Watch Motion Sensor Dataset
   • Accelerometer windowing, feature extraction, manual AdaBoost vs sklearn AdaBoost

---

🧠 Learning Outcomes

• Hands-on implementation of ML algorithms (regression, boosting, SVM, Naïve Bayes)
• Understanding preprocessing, regularization & model evaluation
• Experience with sensor, medical & text datasets
• Practical ML pipeline skills (EDA → preprocessing → model → metrics)
• Data scraping & automation using Python

---

🧰 Tools & Libraries Used

• Python 3.x
• NumPy, Pandas
• Scikit-learn
• Matplotlib & Seaborn
• BeautifulSoup, Selenium
• Jupyter / Spyder IDE

---

✍️ Author

Mehak
B.Tech – Computer Engineering (3rd Year)
Thapar Institute of Engineering and Technology
📧 mmehak2_be23@thapar.edu

---

⭐ Acknowledgement

This repository is part of the Machine Learning (UML501) coursework under the guidance of faculty at TIET, Patiala.

---