Social network analysis for truth Discovery and intelligent decision Making

Abstract

When aggregating information from different sources, one of newlinethe most challenging and essential tasks is obtaining trustworthy facts newlinefrom social media networks like Facebook, Twitter, and Instagram. Post newlinecovid the usage of social networks and generation of voluminous amounts newlineof data is inadvertent. Also, Insights are generated from this data and newlinedecisions are taken by organizations in various domains. So, the most newlineimportant task at hand is analyzing the spread of misinformation newlinedynamics by automatic truth discovery and fact-finding methods. A newlinehumanitarian perspective on these data sources necessitates evaluating newlinethem considering the spread of false information. newlineThis research thoroughly examined the advantages and newlinedisadvantages of current truth discovery algorithms, and a few newlineframeworks and difficulties were discussed. This research covers various newlinetruth discovery strategies on heterogeneous data types like structured newlinedata, unstructured data, and crowdsourced data. Improvements and newlinestrategies are proposed for the entire pipeline of the truth discovery newlineprocess starting from Information extraction and loading till the newlinevalidation against ground truth. newlineKeeping in view the application we are designing privacy and newlineconfidentiality of the data has also been addressed using privacy newlinepreserving truth discovery techniques. It is interesting to see the precision newlineand recall results considering the condition Expectation Maximization newlinealgorithm with Differential privacy and the Expectation Maximization newlinealgorithm without Differential privacy, the difference between the two newlineix newlinewith regards to recall is 1.29% increase and with regards to precision is a newline0.2% decrease. Thus, the impact of privacy preservation on precision and newlinerecall scores has been studied thoroughly. Then, effective feature newlineextraction technique has been proposed which will extract the top 10 newlineessential features from a crowd sourced dataset for better classification newlineaccuracy. Further, in this truth discovery life cycle process a novel truth newlineprediction algorithm