Fake News Predictor App
Co-authors: Vu Luong & Viet Hoang Tran Duong
Introduction
In the following sequence of Articles, we are going to build a web app that predicts if an article is fake or real given the article’s title, author, and body. The app is built on a trained ML model’s API that we will create from the ground up.
Data
The dataset was posted by UTK Machine Learning Club three years ago. The Train.csv file which our team used included five columns. The first column was the id, which was used as an index column. The second column was Title, which included the title of the article, including the publisher. The third column Author included the name of the author, while the last feature column was the Text, which was the main body of the article. The text of the articles wasn’t necessarily complete. The outcome resided in the label column, where 1 meant unreliable/fake news, while 0 meant a reliable article.
import pandas as pd
data = pd.read_csv('train.csv', index_col='id')
data.head()
data.info()
There are some null values in the dataset, so the dataset needs cleaning.
Articles
III. Three Classification Algorithms Comparison
IV. Classification (transformers)
V. Web App and Overall findings
Web App
http://fakenewspds.pythonanywhere.com/
Data & Notebooks
https://drive.google.com/drive/folders/0AJKXYo9Oc9i9Uk9PVA
