Skip navigation
Please use this identifier to cite or link to this item: http://arks.princeton.edu/ark:/88435/dsp015d86p338n
Full metadata record
DC FieldValueLanguage
dc.contributor.advisorMulvey, John M
dc.contributor.authorLi, Xiaoyue
dc.contributor.otherOperations Research and Financial Engineering Department
dc.date.accessioned2022-05-04T15:29:46Z-
dc.date.available2022-05-04T15:29:46Z-
dc.date.created2022-01-01
dc.date.issued2022
dc.identifier.urihttp://arks.princeton.edu/ark:/88435/dsp015d86p338n-
dc.description.abstractPortfolio management is among the most important problems in quantitative finance, with large audience from individuals to institutional investors. Multi-period financial models provide superior capabilities over myopic counterparts, but, in general, suffer from the curse of dimensionality. In this thesis, we examine and solve three multi- period financial models with the assist of modern approaches. First, we adopt a model predictive control approach to solve multi-period portfolio problems with either mean-variance or risk-parity objective functions. The framework provides transaction control by imposing a transaction penalty in the objective. For the risk-parity objective, we provide a successive convex program algorithm with faster and more robust solutions. Out-of-sample tests show promising returns on various asset universes compared to benchmark portfolios. Then, we propose a combined algorithm of dynamic programming with a recur- rent neural network. The dynamic program provides advanced starts for the neural network. Empirical tests show the benefits of this novel strategy with optimizing a portfolio in a regime-switching market in the presence of linear transaction costs. Test problems with 50 to 250 time steps and up to 11 risky assets are solved efficiently, rel- ative to standalone dynamic programs or neural networks. The algorithm addresses the allocation problem in polynomial time as the problem complexity grows. Last, we shift the gear to the problem of optimal execution. In portfolio man- agement, it is not only important to find the optimal allocation, but also how to trade optimally to the desired position. Here, we propose a numerical framework for the optimal portfolio execution problem where multiple market regimes exist. Our approach accepts impact cost functions in generic forms for both temporary and permanent parts. In our numerical experiment, the proposed combined method pro- vides promising selling strategies for both CRRA (constant relative risk aversion) and mean-variance objectives.
dc.format.mimetypeapplication/pdf
dc.language.isoen
dc.publisherPrinceton, NJ : Princeton University
dc.relation.isformatofThe Mudd Manuscript Library retains one bound copy of each dissertation. Search for these copies in the library's main catalog: <a href=http://catalog.princeton.edu>catalog.princeton.edu</a>
dc.subjectdynamic program
dc.subjectmodel predictive control
dc.subjectmulti-period optimization
dc.subjectneural network
dc.subjectportfolio management
dc.subjectportfolio optimization
dc.subject.classificationOperations research
dc.subject.classificationFinance
dc.titlePortfolio Management under Multi-Period Frameworks with Modern Approaches
dc.typeAcademic dissertations (Ph.D.)
pu.date.classyear2022
pu.departmentOperations Research and Financial Engineering
Appears in Collections:Operations Research and Financial Engineering

Files in This Item:
File Description SizeFormat 
Li_princeton_0181D_14012.pdf5.1 MBAdobe PDFView/Download


Items in Dataspace are protected by copyright, with all rights reserved, unless otherwise indicated.