This repository contains the projects for the machine learning class. There are three small projects and one final project.
- Project 1: Using normal perceptron, average perceptron, MIRA, and aggressive MIRA to predicted the salary level based on the provided demographic information. This is a supervised binary classification problem.
- Project 2: Continuing on the salary classification problem, but with SVM. Linear SVM (sklearn), quadratic kernel SVM, and Online SVM (pegasos) are used to conduct the experiment.
- Project 3: Using structured perceptron to implement the speech tagging, which is a common task in Natural Language Processing. Also engineering the feature template to compare the different prediction accuracies.
- Final project: Kaggle competition: Porto Seguro's Safe Driver Prediction. The goal of this project is to predict the probability that if a driver will file a insurance claim or not next year.