Prediksi Trend Harga Saham Perusahaan Tambang Emas dengan Net Foreign Flow menggunakan Algoritma Extreme Learning Machine (ELM) dengan Particle Swarm Optimization (PSO) Berbasis Aplikasi Mobile

Abstract

This study discusses the development of a stock price movement prediction system for gold mining companies based on a mobile application by utilizing the Extreme Learning Machine (ELM) algorithm optimized using Particle Swarm Optimization (PSO). The dataset used consists of stock price data and Net Foreign Flow, which has undergone preprocessing stages and was divided into training, validation, and testing data. Hyperparameter tuning results indicate that applying the sigmoid activation function with 9 hidden neurons produces the best validation performance with an RMSE value of 5.84. Meanwhile, the ELM-PSO combination achieved optimal validation performance with an RMSE of 5.6021. In the testing phase, the ELM model obtained an RMSE value of 15.44, while the ELM-PSO model produced an RMSE of 18.0837. The results of this study show that the ELM algorithm is quite effective in modeling stock price fluctuations, while the PSO optimization improves accuracy during the validation stage.