Fake News Detection Through Feature Weight Optimized Lasso Regression (FWO-LAR)

The massive growth of internet enabled devices increases the users every day active in social media networks. People on social groups getting instant alerts on news, entertainment, education, business and many more. The mode of communication and message transfer through social networking groups have become wider. On the other hand, fake information is widespread over the network through malicious user profiles. These kinds of information change the people perspectives on sensitive matters like social issues, political, religious even more. Detection of fake news and trustworthiness in the received information is important. Considering the major problem of fake news spreading in the social networking groups, an effective system to detect the trustworthiness of the information transferred in the network is considered. The proposed system focused on creating a feature weight optimized lasso regression (FWO-LAR) model for efficient detection of fake news. The system considers the twitter dataset PHEME towards working on goal of detection of fake keywords and fake information associated within the given data. The presented system analyzes the dataset and further evaluate the correlation score, accuracy of detection using FWO-LAR model as 98.7% is achieved with current analysis. The presented study achieved with fast detection of fake news present in the sample data opted under testing. The presented system is compared with existing state of art approaches in terms of performance measure and false positive rate reduction.