Forecasting Saham Perbankan Dengan Penerapan Multilayer Backpropagation Neural Network

I Putu Ryan Paramaditya
Cokorda Rai Adi Pramartha
I Gede Arta Wibawa
I Gede Santi Astawa
Ida Bagus Gede Dwidasmara
I Dewa Made Bayu Atmaja Darmawan

(JELIKU) Jurnal Elektronik Ilmu Komputer Udayana
Vol 12 No 4 (2024): JELIKU Volume 12 No 4, May 2024

The use of the Neural Network algorithm with Backpropagation is used to predict stock price data based on the closing price of the following day, as a reference for buying shares in the future. The dataset used is of the time-series type, stock data for the state-owned banking category comes from Yahoo Finance such as Bank BNI (BBNI). Where from the results of the model training carried out, the lowest loss was 0.0011 at epoch 29, 33, 41, 43, 46, 47, and 49 and the highest was 0.0243 at epoch 0. The lowest Val Loss was 0.0011 at epoch 5, 10, and 46 and the highest was 9.555 at epoch 44. The model test score results showed a Median Absolute Error (MAE) of 85.57 and a Mean Absolute Error Percent (MAE%) of 1.97%. Root Mean Squared Error (RMSE) is 103.85 and Root Mean Squared Error Percent (RMSE%) is 2.39%. This score is considered good because it is below 50%. Prediction results reach an average of above 90%. To get the best prediction results, the percent change must be above -4.35% and the percentage above 95.65%.