Machine Learning: Futures Prognosticator

Abstract

The world composed of different countries has highly emerged nowadays. Foreign currency exchange is the cornerstone of the international market. Therefore, the project aims to study into the data of Forex and forecast the tendency by implementing a Java application. However, forex market prediction is a tedious and complicated task that involves the processing of large amounts of data, which are non-linear and susceptible to economy, politics and other environment factors. Therefore, this project focuses on how advanced artificial intelligence algorithms, like neural networks, applies to forex market prediction and by which means the machine-learning algorithm is implemented into the program. The main challenges of this project are: a) how to combine Pearson correlation algorithm and Levenberg Marquardt algorithm; b) how to set the parameter to the neural network algorithm to better training data and avoid over fitting; c) how to plot figures by using basic java functions without import external drawing jar and d) how to evaluate the system although there is an inevitable error between actual value and predicted data. This project was examined to propose a feasible approach to cope with the fluctuations of foreign currency and also yields good prediction accuracy with preferably mean square error (mse) below 1.22e-03