Clinical Data Analysis and Multilabel Classification for Prediction of Dengue Fever by Tuning Hyperparameter using GridsearchCV

The dengue infection is caused by the mosquito Aedes aegypti. According to WHO 50 to 100 million dengue infections will occur every year in around 80 countries in the world. Data-mining techniques will extract information from the raw data. Dengue symptoms are fever, severe headache, body pain, vomiting, etc. Dengue fever is divided into four classes DEN-1, DEN-2, DEN-3 & DEN-4, and broadly classified as Dengue fever, Dengue Haemorrhagic fever and Dengue Shock Syndrome abbreviated as DF, DHF, and DSS. The research work is carried out on real data and the patient data is collected from the Department of General Medicine, PESIMSR, Kuppam, Andrapradesh. Data was collected for a period of 18 months from January 2020 to June 2021. Dataset consists of 18 attributes and one target value. Research work has been done on binary classification using different ML techniques. The proposed model can be analysed with a multiclass classification approach. The Dengue dataset has divided into 4 levels that are labelled numbers 0-Non dengue, 1-DF, 2-DHF, and 3-DSS. In the proposed model Random Forest with Gridsearch algorithm is used to diagnose the class or level of dengue fever by tuning the hyperparameters with the grid search approach for the prediction. To overcome the problems of the bias of the models 10-fold cross-validation technique is applied. Hyperparameter tuning gives a positive impact on classification accuracy. Performance parameters used for the analysis are accuracy, precision, recall, and f1 score, and compared the proposed model with other ML models. The experimental analysis demonstrates that the RF with GridsearchCV achieved 100% and 98.79% accuracy for training and testing dataset respectively. Comparative analysis shows that the ensemble model gives better precision and recall and f1_score for the multilabel classification of dengue than other ML models.