Implementing classification methods and techniques (ex: Regression Methods, SVMs, Lasso, Supervised Learning, etc.) of Machine Learning to real-world data.
Homework 1: Classification using KNN on Vertebral Column Data Set
Exploring the influence of different K values, distance metrics and voting methods on data sets
Homework 2: Linear Regression on Cycle Power Plant Data Set
Test the results of:
- Simple linear regession for each feature
- Multiple regression
- Adding nonlinear terms and interaction terms to the model.
- Comparison of the results of linear regression with KNN regression
Homework 3: Time Series Classification on Human Activity
Feature extraction from time series:
- Binary Classification with Logistic Regression with Recursive Feature Elimination
- Binary Classification with L1-Penalized Logistic Regression
- Multiclass Classification using Naive Bayes classifier with Gaussian/Multinomial priors
Homework 4:
LASSO and Boosting for Regression on Communities and Crime Rate Prediction
- Ridge/LASSO Regression
- PCR
- Boosting Tree (XGBoost)
APS Failure data Classification using Tree-Based Methods
- Random Forest
- Logistic Model Tree (Weka)
- Compensating Class Imbalance (SMOTE)
Homework 5:
Multi-class and Multi-Label Classification Using Support Vector Machines
Testing the results of a one vs. all classifier to train a SVM for each of the labels with Gaussian kernels and L1-penalized
K-Means Clustering on a Multi-Class and Multi-Label Data Set
- Perform K-Means Clustering on data set and determine majority labeling by reading true labels to calculate the Hamming score
- Monte-Carlo Simulation to retrieve average and standard deviation of running multiple times
Homework 6:
Supervised, Semi-Supervised, and Unsupervised Learning
Examining the results of supervised, unsupervised, and semi-supervised learning methods on the same dataset.
Active Learning Using Support Vector Machines
Implement and compare Passive Learning and Active Learning using SVMs