DSCI 552: Machine Learning for Data Science

Practical applications of machine learning techniques to real-world problems, focusing on data mining, recommendation systems, and adaptive interfaces.

📌 Project Overview

This course involved implementing a wide range of machine learning models from scratch and using industry-standard libraries. The curriculum covered everything from exploratory data analysis (EDA) to advanced deep learning techniques like Transfer Learning.

🛠 Tech Stack

• Languages: Python
• Libraries: Scikit-learn, TensorFlow, Keras, NumPy, Pandas, Matplotlib, Seaborn
• Tools: Jupyter Notebook, OpenCV, Google Colab

✅ Completed Tasks

• KNN Classification: Performed EDA and implemented K-Nearest Neighbors on the Vertebral Column dataset.
• Regression Analysis: Built Linear, Multiple, and KNN Regression models with Hypothesis Testing.
• Time Series Classification: Extracted time-domain features and applied Bootstrapping on AReM data.
• Supervised Learning: Implemented Logistic Regression, Decision Trees (with Cost-complexity Pruning), and SVMs.
• Ensemble Methods: Applied Random Forest, XGBoost, and handled class imbalance using SMOTE.
• Unsupervised Learning: Performed K-Means Clustering and Monte-Carlo Simulations on MFCC data.
• Transfer Learning: Developed a bird species image classifier using EfficientNetB0 and VGG16 architectures.

👨‍🏫 Course Information

• Instructor: Prof. Mohammad Reza Rajati
• Course: DSCI 552: Machine Learning for Data Science, University of Southern California