Abstract
This study develops a stock price prediction model based on the Adaptive Neuro-Fuzzy Inference System integrated with Fuzzy C-Means clustering (ANFIS-FCM) for three information technology stocks listed on the Indonesia Stock Exchange: Anabatic Technologies (ATIC), Elang Mahkota Teknologi (EMTK), and Metrodata Electronics (MTDL). Daily OHLC data from 2015 to 2025 were transformed using a five-day sliding window and normalized with Min-Max scaling. Fuzzy C-Means was used to generate cluster centers for the fuzzy inference system, avoiding the rule explosion problem associated with grid partitioning on twenty input features. Hyperparameters (number of rules, epochs, and learning rate) were tuned separately for each stock using grid search based on validation RMSE. Testing results show MAPE values of 1.95% for ATIC, 3.01% for EMTK, and 1.64% for MTDL, all well below the 10% accuracy threshold commonly used in time-series forecasting. Model performance was found to be data-dependent, with the lowest error obtained for the most stable stock and the highest for the most volatile one.
