A Study on Fake News Detection and Classification Based on Logistic Regression (LR) and Artificial Neural Networks (ANN)

Abstract

In today's digital era, when false information can travel swiftly and sway public opinion, spotting fake news has
become an enormous task. This research uses machine learning algorithms and natural language processing
methods to tackle the issue of false news identification. The goal is to create a trustworthy and precise model that
can determine whether or not a news piece is phoney. Data is loaded, processed, and divided into train and test sets
before classification, prediction, and output production can begin. To do this, we gather and preprocess a labelled
dataset with samples of both false and real news stories, removing noise and extraneous details along the way. A
classification model is trained by dividing the dataset into training and testing sets. The model for detecting false
news is constructed using two classification algorithms: Logistic Regression (LR) and Artificial Neural Networks
(ANN). In contrast to ANN's ability to capture complicated nonlinear connections in the data, LR only gives a linear
decision boundary. Accuracy, precision, recall, and F1 score are only few of the measures used to assess the results
of training both algorithms on the preprocessed data. The findings show that both LR and ANN are very effective in
identifying fabricated stories. LR's interpretability makes it simpler to grasp what criteria led to a classification
being made. When it comes to identifying subtle connections and patterns in the data, ANN excels. The results of
this study could help researchers create better tools to spot bogus news.