Penerapan Jaringan Saraf Tiruan Backpropagation untuk Memprediksi Indeks Harga Saham LQ45

Abstract

Stock price movements are very volatile from time to time. The stock price movement is influenced by many factors, including company performance, dividend risk, the country’s economic conditions, and inflation rate. The existence of these complex factors makes stock price movements challenging to predict. Investors need stock price predictions to see the company’s stock investment prospects in the next period. The method that can predict stock prices is Backpropagation. The Backpropagation method is an algorithm that adopts a human mindset systematically to minimize the error rate by adjusting the weights based on differences in output and the desired target. This study uses historical stock index data for LQ45 from February 26, 2019 – February 26, 2021, namely the closing price as an input and the opening price as the target. The best network model from the Backpropagation method uses a binary sigmoid activation function with nine neurons in the hidden layer. The testing accuracy value is 95.2481% (MAPE), and the error value is 0.000266 (MSE). The error value shows that the prediction model results are excellent