Performance of Correlation in Topic Modeling from Academic Social Network Dataset through Deductive Learning

Large collections of documents are readily available online and widely accessed by diverse
communities. Topic models can extract surprisingly interpretable and useful structure without any explicit
“understanding” of the language by computer. The objective of this work to implement the leading machine
learning algorithms , to get the optimal model through the one of the leading metric Matthew Correlation
Coefficient. This work shows that the accuracies of the NaiveBayesMultinomialText classifier produces
64.59% level of accuracy, IBK classifier is 99.65% level of accuracy,AdaBoostM1classifier is 99.36% and
ZeroR classifier is 64.60% and DecisionStump classifier is 72.22%. The DecisionStump algorithm, Instance
based classifier and AdaBoost Classifiers are correlated positively, but this proposed system recommends
that AdaBoost Classifier and IBK classifers are strongly correlated with this model.