An Ensemble Heterogeneous Hybrid Model for Fake News Detection

Abstract

In fact, fake news has a huge impact on every aspect of our lives, including politics, finance, education, democracy and business. Our aim is to help and provide ability to the users to classify fake real news and avoid the harm of spread unreal news. In our project, we have provided an optimal solution to detect fake news by using an ensemble hybrid model that is a combination of different classifiers: Naive Bayes Classifier, K –Nearest Neighbor classifier, Decision Tree, Bagging Decision Tree, Boosting Decision Tree, Random Forest, Logistic Regression and Support Vector Machine. To implement these classifiers, we have used the Sci-Kit Learn python library. To implement our work, we have used google colab online platform. In our proposed model, for data preprocessing, we have used Natural Language Toolkit (NLTK) technique and for feature extraction, we have used a famous vectorizer called as ‘Term Frequency - Inverse Document Frequency’ (TF-IDF). Then in data segmentation, we’ve splited our dataset into two sets: training set (80%) and testing set (20%). Conclusion: Using the proposed model, the accuracy on the test dataset is 99.13%, with precision of 99.42%, recall of 98.85%, and fl-score of 99.13%.