Intelligence in Finance and Economics for Predicting High-Frequency Data

Forecasting exchange rates is a complex problem that has benefitted from recent advances and research in machine learning. The main goal of this study to design implement method improve the learning performance artificial neural networks with large volumes data using population-based metaheuristics. micro-genetic training algorithm thoroughly analyzed profiling tools find bottlenecks. We compare use predict changes currency on set containing more than 500,000 values. To best parameters networks, we propose an improved by dividing into mini batches. In case, proved be much faster compared standard genetic algorithm, while achieving same prediction accuracy. This allows for just-in-time predictions high frequency data. Here, network models are first created validated existing set. Then, new values can added retrained short time.