FAKE NEWS DETECTION
IN REAL TIME

Constructed an advanced Machine Learning model with 97% accuracy in identifying fake news. • Technologies: Python, Scikit-learn, Flask, Machine Learning, Exploratory Data Analysis, Data Visualization, Preprocessing Data. • Designed and integrated a real-time prediction web application.

fake news detection

Project Description​

This project develops a machine learning model to classify news as real or fake using a rich dataset. It achieves 99.63% accuracy with a Decision Tree classifier, optimized through hyperparameter tuning. A web application enables real-time prediction, aiding efforts to combat fake news.

Data:

Uses a comprehensive dataset consisting of fake and true news articles collected from diverse sources, providing a balanced foundation for training the machine learning models to accurately distinguish between real and fake news.

Preprocessing:

Involves rigorous data cleaning, including the removal of non-alphanumeric characters, conversion to lowercase, tokenization, lemmatization, and stopword removal, ensuring the text data is properly formatted and ready for effective machine learning model training.

Train Models:

Trains multiple machine learning models, including Multinomial Naive Bayes, Logistic Regression, Passive Aggressive, Random Forest, and Decision Tree classifiers, evaluating their performance in accurately classifying news articles as real or fake.

Model Selection:

The Decision Tree (DT) classifier is selected as the best-performing model, achieving an impressive 99.49% accuracy in classifying news articles, making it the most effective model for the task.

Hyperparameter Tuning:

Further optimizes the Decision Tree classifier using RandomizedSearchCV, improving its performance from 99.49% to 99.63% accuracy, fine-tuning the model for even better results.

Web API Development:

Develops a user-friendly web API that allows users to input news articles and receive real-time feedback on their veracity, making the powerful machine learning model accessible to the public.

Deployment:

Successfully deploys the web application, making it available for users to interact with, enabling real-time verification of news articles and contributing to the fight against fake news.

Discussion

The Decision Tree model demonstrates strong potential in detecting fake news, but further advancements could be made by exploring ensemble methods, sophisticated feature extraction, and advanced NLP techniques. Ethical considerations ensured data privacy, using only publicly available datasets. However, limitations include potential biases, gaps in training data, and the subjective nature of defining "fake news," which may lead to discrepancies in real-world application. Despite these challenges, the Decision Tree classifier is a vital tool in combating misinformation, providing clarity in an era where distinguishing truth from fiction is increasingly complex.
Scroll to Top