Penerapan Metode Fuzzy Backpropagation Neural Network dalam Peramalan Saham PT. Bank Rakyat Indonesia (Persero) Tbk. (IDX: BBRI)

Abstract

The Fuzzy Backpropagation Neural Network (FBPNN) is a forecasting approach based on neural networks that applies the backpropagation learning algorithm, where input and output data are expressed in fuzzy membership values. The construction of the FBPNN model consists of three key phases: fuzzification, neural network training, and defuzzification. In the defuzzification stage, the Smallest of Maximum (SOM) technique is used to determine the lowest value among the members with the highest degree of membership. This research employs FBPNN to forecast the weekly closing prices of PT Bank Rakyat Indonesia (Persero) Tbk. (IDX: BBRI). The dataset is divided into 90% for training and 10% for testing. The results indicate that the best network architecture is 20–4–2, producing a Root Mean Square Error (RMSE) of 728,5308 for training data and 178,2089 for testing data. The Mean Absolute Percentage Error (MAPE) reached 10,3620% (classified as good) in training and 3,9189% (classified as very good) in testing. The developed model was then applied to forecast the next five periods, showing a tendency of increasing stock prices