Optimization of Long Term Quantitative Market Predictions

Financial markets today are monitored and controlled by machine learning algorithms. The primary objective of this project is to further develop the algorithm for financial market analysis and prediction that the partner possesses at the moment. The algorithm currently demonstrates high accuracy, subject to certain constraints, among which: a small time interval between a prediction and the actual event and not highly efficient computation of indicators. In addition, the current algorithm is missing any form of analysis of the dynamics of distances to training clusters. These drawbacks are proposed to be eliminated. Additional set of market events will trigger computation of prediction to increase the time span between a prediction and the predicted event. Some of the computation will be done in parallel, potentially using a GPU, since the current model can be easily parallelized. To estimate the relationship between the distances to the training clusters and its dynamics, functional analysis will be deployed.

Vadim Mazalov
Faculty Supervisor: 
Dr. Stephen M. Watt