Optimizing Hyperparameters and Performance Analysis of LSTM Model in Detecting Fake News on Social media

Abstract

Fake news detection recently received a lot of attention from the scientific community and demands an optimal solution with high efficiency. Several studies were conducted using unsupervised and supervised learning techniques to address the fake news identification problem. These studies, on the other hand, have some limitations like inefficient model design, improper pre-processing, and poor accuracy. Some factors contributing to poor accuracy include irrelevant features, improper model parameters, imbalanced datasets, and so on. This work proposes an optimized deep learning model for detecting fake news. The developed model uses Long Short-Term Memory (LSTM) to classify fake and real news and utilizes hyperparameter tuning methods such as grid search and random search to customize the hyperparameters of the model. The experimental results indicate that the optimized LSTM model yields 99.65% accuracy using the ISOT dataset and 45.23% accuracy using the LIAR dataset.