Combination of Fuzzy C-Means, Xie-Beni Index, and Backpropagation Neural Network for Better Forecasting Result

Abstract

Accuracy is one of the performance parameters of a method. This research proposes a combination of Fuzzy C-Means (FCM) method with the Backpropagation (BP) method to improve forecasting performance in terms of accuracy. BP algorithm is a supervised learning algorithm which is have good performance for pattern recognition. In some researches, FCM is more efficient and clustering results are better than other methods. However, FCM has a disadvantage that clustering results are affected by clustering configurations, such as the number of clusters. Therefore it is necessary to do cluster validation. One of popular cluster validation method is Xie-Beni (XB) index. In this paper, we propose a forecasting system by combining the validated FCM algorithm using the XB index method with the BP algorithm. The data are grouped using FCM with number of clusters 3, 4, 5, 6, 7, 8, 9, and 10. Then, the clustering results validated using XB and find the most suited number of clusters for the data. Each cluster becomes the input of the BP neural network for forecasting process. This research uses sales data of 49 types of products for 25 months.