TitleDynamic modeling and forecasting algorithms for financial data systems
NameMahajan, Manish (author), Puri, Narindra N. (chair), Gajic, Zoran (internal member), Marsic, Ivan (internal member), Lee, Cheng-Few (outside member), Jaggi, Bikki (outside member), Rutgers University, Graduate School - New Brunswick,
SubjectElectrical and Computer Engineering,
Recursive functions--Data processing,
DescriptionIt is a valid question that why a Control Systems Engineer would be interested in dealing with financial instruments. Financial instruments involving option theory are very elegant, math oriented and practical. These mathematical tools have created a new industry known as 'Derivative Industry' or 'Hedge-Fund Industry' or so called 'Risk-Management Industry'. This thesis is aimed at developing investment strategies involving the decision making needs via control system techniques. The problem, in general, is computationally challenging particularly when investment of many securities is involved resulting in a high dimensional computational framework. Furthermore, complications may arise due to realistic restrictions and non-linearities. The various areas of financial engineering are very fertile for the application of the system methodology and control theory techniques. Modeling, optimization, identification and computational methods used in the Systems Engineering can be successfully applied to the financial instruments. The ideas developed in this thesis are more about the scientific reasoning involving financial instruments rather than specific situations alone. Major contribution of this thesis is the time series optimal prediction filter and the development of the Dynamic Modeling and Forecasting Algorithm (DMFA). The proposed algorithm predicts the next data point of the financial time series while dynamically computing the parameters from existing data. The computation of the parameters is optimized by use of the recursive matrix inversion algorithm. The system is solved via an innovative technique of inversion such that it avoids explicit inversion of more than a 2 X 2 matrix and computation of higher dimensional determinants and co-factors. This results in new contributions to computation finance and numerical methodology along with arbitrage decision and hedging strategies under market uncertainties as well as robust control applications. The minimum mean-square algorithm used assures system stability via poles within the unit circle. The DMFA method is a superior auto regression (AR) model as a general system of time-series realizations in-order to calculate the coefficients that fit the model for a better prediction. Theoretical modeling and market specific volatility models, updated volatility computation are derived from the observation data.
NoteIncludes bibliographical references
Noteby Manish Mahajan
CollectionGraduate School - New Brunswick Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.