daklekar / hw4

HW4 Dhruvakumar Aklekar (801325204)

Problem 1 (50pts):

Use the cancer dataset to build an SVM classifier to classify the type of cancer (Malignant vs. benign). Use the PCA feature extraction for your training. Perform N number of independent training (N=1, …, K).

Identify the optimum number of K, principal components that achieve the highest classification accuracy. Plot your classification accuracy, precision, and recall over a different number of Ks. Explore different kernel tricks to capture non-linearities within your data. Plot the results and compare the accuracies for different kernels. Compare your results against the logistic regression that you have done in homework 3. Make sure to explain and elaborate your results.

Problem 2 (50pts):

Develop a SVR regression model that predicts housing price based on the following input variables:

Area, bedrooms, bathrooms, stories, mainroad, guestroom, basement, hotwaterheating, airconditioning, parking, prefarea

Plot your regression model for SVR similar to the sample code provided on Canvas. Compare your results against linear regression with regularization loss that you already did in homework1. Use the PCA feature extraction for your training. Perform N number of independent training (N=1, …, K). Identify the optimum number of K, principal components that achieve the highest regression accuracy. Explore different kernel tricks to capture non-linearities within your data. Plot the results and compare the accuracies for different kernels.