COMPARING NAIVE BAYES, K-NEAREST NEIGHBOR, AND NEURAL NETWORK CLASSIFICATION METHODS OF SEAT LOAD FACTOR IN LOMBOK OUTBOUND FLIGHTS

Abstract

Classification is the process of grouping data based on observed variables to predict new data whose class is unknown. There are some classification methods, such as Naïve Bayes, K-Nearest Neighbor and Neural Network. Naïve Bayes classifies based on the probability value of the existing properties. K-Nearest Neighbor classifies based on the character of its nearest neighbor, where the number of neighbors=k, while Neural Network classifies based on human neural networks. This study will compare three classification methods for Seat Load Factor, which is the percentage of aircraft load, and also a measure in determining the profit of airline.. Affecting factors are the number of passengers, ticket prices, flight routes, and flight times. Based on the analysis with 47 data, it is known that the system of Naïve Bayes method has misclassifies in 14 data, so the accuracy rate is 70%. The system of K-Nearest Neighbor method with k=5 has misclassifies in 5 data, so the accuracy rate is 89%, and the Neural Network system has misclassifies in 10 data with accuracy rate 78%. The method with highest accuracy rate is the best method that will be used, which in this case is K-Nearest Neighbor method with success of classification system is 42 data, including 14 low, 10 medium, and 18 high value. Based on the best method, predictions can be made using new data, for example the new data consists of Bali flight routes (2), flight times in afternoon (2), estimate of passenger numbers is 140 people, and ticket prices is Rp.700,000. By using the K-Nearest Neighbor method, Seat Load Factor prediction is high or at intervals of 80% -100%.