PROJECT TITLE :

Challenges in KNN Classification

ABSTRACT:

One of the most common and widely used data mining algorithms is called the KNN algorithm. It has been widely applied to data analysis applications across a wide range of research areas in the field of computer science, where it has met with great success. This paper demonstrates that, despite its success, the KNN classification method still faces many challenges, such as the computation of K, the selection of the nearest neighbor, the search for the nearest neighbor, and the establishment of classification rules. After establishing these issues, recent approaches to their resolution are investigated in greater detail. This provides a potential road map for ongoing research related to KNN, in addition to some new classification rules regarding how to address the problem of training sample imbalance. Experiments were run using 15 benchmark datasets from the UCI in order to evaluate the different approaches that were proposed.