Machine Learning

PCA-Logistic Regression For Fake News Detection

The PCa-Logistic Regression google document file is used to determine whether a headline will indicate if the article is from a fake or real website. I have previously conducted similar classification tests using the Naive Bayes and K-Nearest Neighbour approach. In this document I have determined whether an article is real or fake by using the logistic regression algorithm. Furthermore, we will be using the Logistic regression algorithm in combination with:

• an L1 Penalty
• an L2 Penalty
• no Penalty
• a 100-Principal Component analysis reduction
• a 100-Principal Component analysis reduction
• Cross Validation with:
  • an L1 Penalty
  • an L2 Penalty
  • no Penalty

KNN & Naives Bayes

This folder uses KNN & Naives Bayes Classifiers to train and distinguish between real and fake news, based on the article's headline. This code will work with any headline, but it will need to be formatted to work with a custom headline. To run the code just pull the folder from Github and import the dependencies. The KNN and Naive Bayes Classifer will show the Precision, Recall and F1 Scores based on a test(titled Dev) dataset. The two datasets (Train, Dev(test)) are scraped from the web and are broken up evenly into their respective folders. Each dataset will comprise two seperate Fake and Real headlines so the algorithms can be trained and tested on both headlines. The link to the Google Document file is here: KNN & Naives Bayes google Document

PCA-Guassian

This file is used to demonstrate my ability to calculate a Gaussian Model from scratch. In addition, I also conducted a PCA analysis on some data. The link to the Google Document file is here: PCA-Guassian Google Document

This is dummy line