Classification and analysis for Focused Crawled Textual Dataset for retrieving Indian origin scientists

Abstract

Text classification also called (text categorization or text tagging) is a crucial and extensively used approach in Natural Language Processing (NLP), to predict unseen content documents into prearranged categories. In this paper, we evaluate the dataset construction and evaluation process as a component of text classification. To begin with, we produced a newly created dataset for Indian Origin Scientists for text classification, which was collected by applying focused crawling and web scraping techniques. We then demonstrate an extensive evaluation of numerous models on this recently constructed dataset. Our evaluations display that the Random forest model outperforms the rest of the supervised models. Our results produce a fine beginning for additional research in Indian Origin Scientists' classification of text. Investigational outcome with K Nearest Neighbor, Logistic Regression, and Support Vector Machine for Indian-origin scientists produced much better performances for Random Forest when combined with SMOTE and K fold cross-validation techniques. We apply the Area under the ROC Curve to compute the effectiveness of the chosen models. Overall, the Random Forest classifier exhibited the best output along with 90% micro-average AUC.