ISSN (O) 2393-8021, ISSN (P) 2394-1588 IARJSET International Advanced Research Journal in Science, Engineering and Technology Vol. 8, Issue 8, August 2021 DOI: 10.17148/IARJSET.2021.8816 © IARJSET This work is licensed under a Creative Commons Attribution 4.0 International License 69 An Enhanced Random Forest Model for Detecting Effects on Organs after Recovering from Dengue Avijit Kumar Chaudhuri 1 , Arkadip Ray 2 , Prof. Dilip K. Banerjee 3 , Dr. Anirban Das 4 1 Research Scholar, Department of Computer Application, SEACOM SKILLS UNIVERSITY, Kendradangal, Bolpur, Birbhum, 731236, West Bengal, India 2 Department of Information Technology, Government College of Engineering and Ceramic Technology, Kolkata, West Bengal, 700010, India 3 Professor, Department of Computer Application, SEACOM SKILLS UNIVERSITY, Kendradangal, Bolpur, Birbhum, 731236, West Bengal, India 4 University of Engineering & Management, Kolkata, West Bengal, India Abstract : One of the most important issues in healthcare and machine learning research is determining the likelihood of organ failure in dengue fever (DF). Researchers have demonstrated the effectiveness of artificial intelligence and machine-learning models in a variety of practical classification problems. The current study focuses on the development of a prediction model for organ failure in dengue. This paper proposes an Enhanced random forest (ERF) model that employs an ensemble of classification methods to achieve this goal. The proposed ERF classifier is tested on a dengue fever dataset collected from dengue patients from all over the West Bengal state in India from 2016 to 2019, from several hospitals, and by interacting with people previously infected with DF individually using online and offline questionnaire methods. The proposed classifier is also compared to some cutting-edge machine-learning classifiers, including random forest, naive Bayes, support vector machine with radial basis function kernel, and decision tree. To assess the strength of the proposed ERF classifier, various performance metrics such as accuracy, sensitivity, specificity, receiver operating characteristic, area under the curve, and some statistical tests such as kappa statistics were used to test the classifiers. To test the credibility of the classification models in dealing with unbalanced data, various splits of training and testing data – namely, 50–50 percent, 66–34 percent, 80–20 percent, and 10-fold cross- validation – were used in this study. The output results were also compared to previous research on the same dataset, where the proposed classifiers were found to be the best across all performance dimensions. Keywords: Enhanced Random Forest (ERF), Dengue Fever (DF), Organ Failure, Ensemble of Classification Methods, Kappa Statistic, ROC-AUC 1. INTRODUCTION Human well-being is facing significant challenges as instances of various virus-related diseases are increasing. In these situations, focused research and development are a pressing need. Given the availability of broad datasets, data mining is a recommended approach to gain insights. The occurrence of multiple organ dysfunctions is due to the development of mortality in acute dengue infections. Severe dengue symptoms vary, and consequently, the precise morbidity and mortality in terms of organ failure are not well studied in the Indian environment. From 2019 to 2021, the authors conducted a prospective multicenter retrospective analysis in some territory care and intensive care units in Kolkata. The aim was to calculate the likelihood of organ failure in serious dengue infection [11]. The availability of pathological test data offers the potential to gain insights with a higher accuracy level. Some authors have used data- mining approaches in the study of dengue but have concluded that these approaches have varied accuracy in prediction. There is very little evidence in studies on the application of data mining to predict the effects of dengue on several organs. Bagging and boosting are two ways to improve the classification. Bagging performs a bootstrap to obtain object samples to train a classifier in dealing with individual samples [3]. Although the classifier models used in bagging are sensitive to minor data changes, bootstrap sampling appears to lead to low-diversity ensembles compared to other