Diagnosis Of Dengue Fever Using Machine Learning Algorithms

Abstract

The study offers crucial information to academics and medical experts, directing the choice of the best modeling algorithms for infectious illnesses. The whole outcomes of my "Diagnosis of Dengue Fever Using Machine Learning Algorithms" proposal may be seen here. This article goes into detail about how the idea was transformed into a thesis.The objective of this project was to create a system that can accurately diagnosis a dengue using machine learning algorithms , millions of people world wide are infected with dengue fever every year, another virus carried by mosquitoes. Dengue outbreaks are relentless,control strategies must be innovative and aggressive. In this context, Machine Learning (ML) techniques appear to be a promising avenue for increasing our ability to predict the occurrence and spread of dengue fever. This study examines how well different machine learning techniques predict dengue fever using a well compiled dataset of 1,037 items and 12 attribute. Some assembly algorithms were used in this work: XGBoost, Random Forest, AdaBoost and CatBoost & Decision Tree, Naive Bayes, K-Nearest Neighbors(KNN), Support Vector Machine (SVM) models were used. CatBoost outperforms other methods studied with an amazing accuracy of 96%. This accuracy is a testament to the algorithm's ability to learn complex relationships in multidemonsial data sets, making it an excellent candidate for dengue fever diagnosis. The present study emphasizes the early diagnosis of dengue infection using machine learning techniques applied to a huge clinical dataset comprising 1,037 patient records and several hematological and biochemical characteristics. The major goal of this study was to develop and test a reliable automated model for identifying patients as Dengue Positive or Dengue Negative based on blood test parameters such as Platelet Count, WBC, RBC, HCT, Lymphocyte (%), and Neutrophil (%). We testted and compared eight suppervised machine learning algorithms: CatBoost, XGBoost, Random Forest, Decision Tree, AdaBoost, K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Naive Bayes. Each moddel was trrained and then evalluated on preprocessed data in order to juddge the moddels on important mettrics such as accuracy, precision, recall, and the F1-score. The testing results shoowed that the higheest accuracy was by CatBoost at 96.15%, followed by XGBoost with 95.19%, and Random Forest at 94.23%, refllecting that the perfformance of ensemble bassed algorithms is much better in handdling compplicated and nonllinear data patterns. The Decision Tree had an accuracy of 88.94%, while simppler models like Naive Bayes, KNN, and SVM had lower accuracies in here.