Algoritma K-Nearest Neighbors dan Synthetic Minority Oversampling Technique dalam Prediksi Pemesanan Tiket Pesawat

Abstract

This study applies the Synthetic Minority Oversampling Technique to improve the performance of the K-Nearest Neighbors method in predicting the unbalanced data class. Most classification algorithms implicitly assume that the processed data has a balanced distribution, so that the standard classifier is more inclined towards data with a dominant class number (majority class). The use of Synthetic Minority Oversampling Technique can improve the performance of the K-Nearest Neighbors method for flight ticket booking data. Although in terms of accuracy, Synthetic Minority Oversampling Technique with K-Nearest Neighbors is lower at 79.65% compared to K-Nearest Neighbors without using Synthetic Minority Oversampling Technique, which is 97.81%, the suggested technique did not improve but from other performance, The proposed method can outperform K-Nearest Neighbors by using Synthetic Minority Oversampling Technique in terms of precision, recall, and F1-Score when applied to the Airline Ticket Booking dataset. Precision increased 18.00% from 62.00% to 80.00%, recall increased 28.00% from 52.00% to 80.00%, and F1-Score increased 27.00% from 53.00% to 80 ,00% on the flight ticket booking dataset.

Decision Tree C4.5 Performance Improvement using Synthetic Minority Oversampling Technique (SMOTE) and K-Nearest Neighbor for Debtor Eligibility Evaluation

“…SMOTE randomly selects the minority data, then the KNN value is set by the user. Synthetic data is generated between the random data and the KNN value [21]. The final result at this stage is the formation of a new training dataset with balanced class distribution conditions, namely 490 class 1 data and 490 class 2 data, so that the total training data is 980 data.…”