multivariate next-day price prediction ROAMA dataset

the dataset is related to an app-based framework for multivariate next-day price prediction usingGRU attention networks with rolling averages.The major elements of all stock markets are stocks, currencies, commodities, and cryptocurrencies, and predicting the prices of these elements is crucial for nationaleconomies. Among all commodities including silver, platinum, and copper, gold is the one that determines the price movements of the others. This paper proposesROAMA, a unique rolling-average multivariate attention methodology for predicting gold prices by using GRU attention models and simple, cumulative, and exponentialmoving averages as features. ROAMA predicts next-day close, open, high, and low prices based on 63 features. Daily gold price data from December 2011 to September2024 were used in extensive experiments to predict future prices, and the results confirm that ROAMA has superior performance in terms of the highest predictiveaccuracy for close (0.99875), open (0.99899), high (0.99902), and low (0.99895) prices and the lowest corresponding RMSE and MAE values. We developed a simple webapp that reads yesterday’s four prices and predicts today’s prices. Our developed web app can be used by retail investors to obtain possible prices and then makeinformed decisions accordingly.