Improvement Of Prediction Model Using K-Nearest Neighbors (Knn) And K-Means In Medical Data

Lestari, Wiji
Sumarlinda, Sri
Binti Rahmat, Azizah

Proceeding of International Conference Health, Science And Technology (ICOHETECH)
2024: Proceeding of the 5th International Conference Health, Science And Technology (ICOHETECH)

Improving the performance of a prediction model is very important in its implementation. This study aims to improve the performance of the K-Nearest Neighbors (KNN) classification model with the K-Means clustering algorithm. The dataset used is UCI global data with 300 data and 12 features. The dataset is divided into 200 training data and 100 testing data. The training data is then processed by clustering with K-Means. The cluster centroid from the clustering results will be calculated for its distance from the testing data and produce data classification. The results of the classification process show that the accuracy of the proposed model is 76.45% better when compared to the results of the KNN classification process, for k = 5 the accuracy is 63.37%, k = 10 the accuracy is 64.36% and k = 15 the accuracy is also 64.36%.