## USING STOCHASTIC OPTIMIZATION METHODS FOR STOCK SELLING DECISION MAKING AND OPTION PRICING: NUMERICS AND BIAS AND VARIANCE DEPENDENT CONVERGENCE RATES

### BibTeX

@MISC{Bao_usingstochastic,

author = {J. Bao and A. Belu and Y. Gershon and Y. J. Liu and G. Yin and Q. Zhang},

title = {USING STOCHASTIC OPTIMIZATION METHODS FOR STOCK SELLING DECISION MAKING AND OPTION PRICING: NUMERICS AND BIAS AND VARIANCE DEPENDENT CONVERGENCE RATES},

year = {}

}

### OpenURL

### Abstract

Abstract. This paper is concerned with using stochastic approximation and optimization meth-ods for stock liquidation decision making and option pricing. For stock liquidation problem, we present a class of stochastic recursive algorithms, and make comparisons of performances using stochastic approximation methods and that of certain commonly used heuristic methods, such as moving averaging method and moving maximum method. Stocks listed in NASDAQ are used for making the comparisons. For option pricing, we design stochastic optimization algorithms and present numerical experiments using data derived from Berkeley Options Data Base. An important problem in these studies concerns the rate of convergence taking into consideration of bias and noise variance. In an effort to ascertain the convergence rates incorporating the computational efforts, we use a Liapunov function approach to obtain the desired convergence rates. Variants of the algorithms are also suggested. 1. Introduction. The

### Citations

639 |
A stochastic approximation method
- Robbins, Monro
- 1951
(Show Context)
Citation Context ...n approach to obtain the desired convergence rates. Variants of the algorithms are also suggested. 1. Introduction. The original motivation of stochastic approximation introduced by Robbins and Monro =-=[13]-=- is concerned with finding roots of a continuous function f(·), where either the precise form of the function is unknown, or it is too complicated to compute; the experimenter is able to take “noisy” ... |

441 |
Monte Carlo Methods in Financial Engineering
- Glasserman
- 2003
(Show Context)
Citation Context ...tically interesting results, the computation required often deemed to be infeasible in real applications. Thus, taking such practical concern into consideration, Monte Carlo methods were suggested in =-=[4, 10]-=-, and references therein. Nevertheless, the computation required is normally rather extensive; the number of simulation runs is often large (e.g., from 50000 to 100000 or even more). Designing more ef... |

400 | Valuing american option by simulation: A simple leastsquared approach. The Review of Financial Studies
- Longstaff, Schwartz
- 2000
(Show Context)
Citation Context ...tically interesting results, the computation required often deemed to be infeasible in real applications. Thus, taking such practical concern into consideration, Monte Carlo methods were suggested in =-=[4, 10]-=-, and references therein. Nevertheless, the computation required is normally rather extensive; the number of simulation runs is often large (e.g., from 50000 to 100000 or even more). Designing more ef... |

242 |
Stochastic Approximation and Recursive Algorithms and Applications
- Kushner, Yin
- 2003
(Show Context)
Citation Context ...ory, manufacturing and production planning, adaptive control, signal processing, and wireless communications. A most up-to-date development on stochastic approximation can be found in the recent book =-=[7]-=- and references therein. This work continues our effort in using stochastic approximation and optimization methods to make decisions for financial engineering applications. Specifically, we focus our ... |

220 | E¢ cient analytic approximation of american option values
- Barone-Adesi, Whaley
- 1987
(Show Context)
Citation Context ...to demonstrate the effectiveness of our algorithms. Pricing options has been a major research topic in financial engineering for years. There is a substantial research devoted to it; see for example, =-=[1, 2, 5]-=- and references therein. As is well known, pricing American put options can be formulated as an optimal stopping problem. A typical method of solution is a variational or quasivariational inequality a... |

170 |
Stochastic estimation of the maximum of a regression function
- Kiefer, Wolfowitz
- 1952
(Show Context)
Citation Context ...the level of dosage and f(x) is the probability of success (leading to the recovery of the patient) at dosage level x. The classical Kiefer–Wolfowitz (KW) algorithm introduced by Kiefer and Wolfowitz =-=[6]-=- concerns the minimization of a real-valued function using only noisy functional measurements. Tremendous progress has been made in the study of stochastic approximation methods for the past a half of... |

104 | Stochastic calculus for fractional Brownian motion
- Duncan, Hu, et al.
- 2000
(Show Context)
Citation Context ...while undertaking is to consider situations that only part of the stock shares is sold and develop the corresponding stochastic approximation algorithms. Recent advances in fractional Brownian motion =-=[3]-=- open up new research avenues; designing the associated stochastic optimization algorithms deserves further study and investigation. Acknowledgement. We are very grateful to Dr. Reza Kamaly and Necess... |

46 |
Dynamic Asset Pricing Theory, 2nd edition
- Duffie
- 1996
(Show Context)
Citation Context ...to demonstrate the effectiveness of our algorithms. Pricing options has been a major research topic in financial engineering for years. There is a substantial research devoted to it; see for example, =-=[1, 2, 5]-=- and references therein. As is well known, pricing American put options can be formulated as an optimal stopping problem. A typical method of solution is a variational or quasivariational inequality a... |

28 | Budget-dependent convergence rate for stochastic approximation - L’Ecuyer, Yin - 1998 |

28 |
trading: An optimal selling rule
- Zhang, Stock
(Show Context)
Citation Context ...for example, is: When should we exercise an American put option? Since it is a decision making process involving randomness and uncertainty, stochastic control techniques naturally come into play. In =-=[16]-=-, liquidation of a stock was formulated as an optimization problem. Assuming that the stock price is represented by a regime-switching geometric Brownian motion model, a diffusion process modulated by... |

14 |
Recursive algorithms for stock Liquidation: A stochastic optimization approach
- Yin, Liu, et al.
(Show Context)
Citation Context ...ock splits and dividend payouts). Horizontal axis–days; Vertical axis–stock prices mathematical question. Focusing on threshold-dependent decisions, a stochastic approximation method was developed in =-=[14]-=-. A class of easily implementable recursive algorithms was proposed; convergence of the algorithms was obtained via the limit ordinary differential equation of the interpolated sequence of the iterate... |

8 |
How to Make Money
- O’Neil
(Show Context)
Citation Context ... the stochastic approximation method provides much better return. 4.2. Stock Liquidation: Comparisons of Stochastic Optimization and Moving Maximum Methods. This part is inspired by an idea of O’Neil =-=[11]-=-. Starting from a certain day, we check the price of each stock, and make a purchase decision if it is greater than its maximal price of the last 20 days. Once the stock is bought, we use both stochas... |

5 |
liquidation via stochastic approximation using NASDAQ daily and intra-day data
- Yin, Zhang, et al.
(Show Context)
Citation Context ... iterates; rates of convergence were ascertained using stationary covariance of associated diffusions together with scaling factors. Simulations and real data were also demonstrated. Subsequently, in =-=[15]-=-, probability estimates of the iterates escaping from a neighborhood of the optimal threshold value were derived. Moreover, we used issues from NASDAQ 100 daily closes and intra-day data to demonstrat... |

1 |
Computational methods for pricing American put options
- Liu, Yin, et al.
(Show Context)
Citation Context ...ength needed to take θn + εn∆� φn back to the constraint set [θl, θu] if it ever escapes from there. Under broad conditions, it can be shown that the above algorithm converges. We refer the reader to =-=[9]-=- for further details. We reiterate that unlike the Monte Carlo approach, the precise model need not be known. The recursion depends on the observed data only. More will be said in the numerical experi... |

1 |
Robust pseudogradient adaptation algorithm
- Tsypkin
(Show Context)
Citation Context ...is a sufficiently large positive number. Note that in this case, the mean limit ordinary differential equation becomes ˙θ = ∇φ(θ) − K0∇c(θ). 5.2. Robust Algorithm. The motivation comes from a work of =-=[12]-=-. Since the actual dynamics might be hardly known, one may use the idea in robust statisticals130 J. BAO ET AL. analysis. Let � Bi(·), i = 1, 2 be bounded real-valued functions. In lieu of (4), consid... |