Fuzzy Integration to Standard Calculation of K-Nearest Neighbour Attributes

Abstract

The development of information and data in the era of the industrial revolution 4.0 is very fast. Researchers, institutions and even industry are competing to find and utilize methods in data processing that are more effective and efficient. In data mining classification, there are several best methods and are widely used by researchers. One of them is K-Nearest Neighbor (KNN). The calculation process in the KNN algorithm is carried out by comparing the testing data to all existing training data. This comparison is generally symbolized by the value of closeness or similarity between attribute records. The KNN method is proven to be good for handling large datasets and datasets with many attributes. One of the drawbacks in calculating the similarity of the KNN is that if there are attributes with a large range value, the similarity value will also be large. Conversely, if the range in an attribute is small, the similarity is also small. This condition is clearly unfair considering the types of attributes in the current data vary widely. One solution to this problem is to use standardization for all existing data attributes. Fuzzy is a model introduced by Prof. Zadeh which allows a faint value to be a value between 1 and 0. In this study the fuzzy model will be integrated in the KNN similarity calculation to obtain standardization of all data attributes. The results show that the use of the KNN algorithm in the classification of credit approval has an accuracy rate of 91.83%.