Machine-learning classification techniques for the analysis and prediction of high-frequency stock direction   رسالة دكتوراه