Boosting Disease Prediction Accuracy: Comparative Evaluation of KNN
Author : Shaheen Khatoon, Aparna Tiwari and Rinku Raheja
Abstract :
These days, machine learning is important in a variety of domains and helps businesses make wise judgments. It is important to note that (ML) is currently a crucial idea to expand artificial intelligence (AI) of smart systems by reducing the need for human interaction due to humans' limited memory and brain capacities. Artificial Intelligence (AI) is a branch of computer science study that tries to replicate human cognitive processes in machines. It has applications in several domains such as industry, commerce, energy, transportation, health, and security. Furthermore, artificial intelligence cleared the way for the development of contemporary technologies that struggle to automate processes and function without human intervention.
One of the greatest classification techniques is the kNN algo, which is a well-known pattern recognition technique. It is among the most straightforward machine learning methods for categorization. This paper explores the use of machine learning algorithms, specifically the k-nearest neighbour (KNN) algorithm, in disease risk prediction. The datasets were related to different disease contexts. For comparative study, we took into account the accuracy, precision, and recall performance metrics. These variations' average accuracy values varied from 64.22% to 83.62%. The ensemble method KNN (82.34%) and Hassanaat KNN (83.62%) both displayed the highest average accuracy. To evaluate each variant and compare the outcomes, a relative performance index is also suggested, depending on each performance indicator. Based on the accuracy-based version of this index, this study found that the Hassanat KNN variant performed the best, followed by the ensemble approach KNN. This paper provides an overview of the kNN algorithm and its related literature, delves into the algorithm's concept, stages for implementation, and implementation code, and evaluates the benefits and drawbacks of the different improvement approaches with better accuracy score 97.3%.
Keywords :
kNN algorithm, k nearest neighbor algorithm, machine learning, blockchain.
