ALGORITMA GENETIKA UNTUK OPTIMASI PARAMETER NILAI K PADA METODE K-NEAREST NEIGHBOR

Abstract

Classification is a systematic grouping of objects into certain categories based on the common characteristics they have. One of the algorithms that is often used in classification is the K-Nearest Neighbor (KNN) algorithm, because the algorithm is easy to understand and apply, can be used on data that has many classes, and is effectively used for large data. However, this algorithm also has a weaknesses in making biased K parameters, resulting in reduced level of accuracy. Therefore, in this study, optimization of K parameters was carried out using a genetic algorithm.The data used in this study is a drug dataset sourced from the Kaggle platform. The total number of these datasets is 200 data records with 150 data used for training data and 50 data used for data testing. The overall data will be classified into five categories, namely drug category A, drug B, drug C, drug X, and drug Y. The classification is based on five criteria, namely gender, age, blood pressure, cholesterol, and sodium potassium. The best accuracy results obtained from the optimized KNN classification is 85% with parameter K=1. The accuracy is the same as the KNN search results without optimization by using the K parameter from a value range of 1 to 5 which produces the best accuracy when the parameter is at K = 1.