Our goal in this paper is to study the behavior of solutions for stochastic differential equations (SDEs) such as the Heston
model (see [1]). The Heston model is a very useful stochastic volatility model used in financial markets where the evolution
for the stock price volatility is described and the volatility is a random process. Market situations at relatively low volatility
levels have attracted academic attention. For example, Duran and Bommarito (see [2]) argued that there may be a temporary
silence at the beginning of a credit crunch, especially when prices are overvalued at a high level, and the low market
volatility may indicate the silence before a storm. We extend our approach (see [3]) by using several numerical solution
methods for the Heston stochastic volatility model and by applying to Borsa Istanbul-100’s (BIST-100) large data set. We
use Euler–Maruyama, Milstein and stochastic Runge–Kutta (SRK) methods (see [4–7]). We compare the trade-off between
cost and robustness of the methods while choosing a suitable method for the application dealing with the relatively low
volatility cases. We perform simulations for different stock market conditions by using the real large data set.
At the application of the Heston model,wealso need to know the initial and long term variances like the other parameters.
However, it can sometimes be hard to estimate these values for a large data set. For this purpose, the natural log ratio of the
highest price to the lowest price (see [8]), standard deviation or simple return based on the highest and the lowest prices
in the interval can be considered by taking their squares. We prefer volatilities having simple return of extreme values at
the overlapping case which is applied for closed-end funds (see [9]) when we examine initial and long term volatilities of
BIST-100 index for each year between 2007 and 2012. We also employ unit volatility of extreme values to estimate the
volatilities of BIST-100 index in our analyses.
It is important to find a method to quantify market impression approximately, and summarize large data set in the
presence of several variables together. Although market price reflects all past publicly available information according to
weak-form efficient-market hypothesis (EMH) (see [10]), many traders believe that prices can be overvalued or undervalued.
Therefore, we seek to find a scheme for market impression in addition to market price. We believe that market impression
may be expressed via several variables such as volatility, interest rate, and time, together.
The remainder of the paper is organized as follows. In Section 2, we introduce 3-dimensional matrix norms as generalizations
of the matrix norms and we prove them by using the applicable numerical linear algebra and analysis arguments
(see [3,11,12] and references therein). In Section 3, we define the market impression matrix norm as an application to the
3-dimensional matrix norms. In Section 4, we use extreme value based volatility at the overlapping case and present results.
Section 5 concludes the paper. Appendix includes Milstein, stochastic Runge–Kutta methods and Heston model.