![](bg2.png)
2 RELATED WORK
Stock market analysis has been one of the most
attractive and active research fields, where many
Machine Learning techniques have been used.
Generally speaking, one can distinguish two
methods for anticipating future stock prices and the
time to buy or sell; one is Technical Analysis
(Murphy, 1999) and the other is Fundamental
Analysis (Graham et al., 2003). Fundamental
Analysis look at stock prices using financial
statement of each company, economic trend and so
on; requires a large set of financial and accounting
data, difficult to obtain and both released with some
delay and often suffers of low consistency.
Technical Analysis numerically analyzes the past
movement of stock prices, is based on the use of
technical stock market indicators that work on a
series of data, usually stock prices or volume,
(Achelis, 2000) is accurate, on time, and relativity
easy to obtain. Consequently, this work will be
focused on the use of Technical Analysis to
anticipate future stock price movements.
Many approaches based on evolutionary
computation have been proposed and applied to
diverse fields of financial to predict worth trends. In
an attempt to summarize, in most of the works, the
generated returns are exclusively used as the only
fitness metric, without accounting for the related
risk. Some examples are the use of GAs to optimize
TI's parameters (Fernández-Blanco et al., 2008), or
to develop TS based on TI's (Bodas-Sagi et al.,
2009), (Gorgulho et al., 2011).
According to what was stated for the first time in
1952 (Markowitz, 1952), any TS should have the
highest possible profit with the feasible minimal
risk. Sadly, these two metrics are intrinsically
conflicting by virtue of the risk-returns trade-off.
Some articles propose the combination of the two
conflicting objectives into one single metric, in
particular (Bodas-Sagi & al., 2009) use the Chicago
Board Options Exchange (CBOE) Volatility Index
(VIX) as an estimate of risk. Also, (Schoreels & al.,
2006) propose the use of a Capital Asset Pricing
Model (CAPM) (William, 1964) system, based on
portfolio theory (Markowitz, 1952) to reduce risk
trough balanced selection of securities. More
recently (Pinto et al., 2011) propose and study
several alternatives to the classical fitness evaluation
functions.
A Multi-Objective system to maximize the total
returns and to minimize the risk as the exposure to it
is proposed by (Chiam et al., 2009). The framework
is tested using data gotten from one stock market,
the Singapore Exchange stock market (Straits Times
Index (STI)). Hence, some of the conclusions drawn
on this study could be attributed to the market used
to test it. Moreover, the metric used to evaluate the
return is peculiarly unusual; so, it is difficult, to
compare the presented results with the results
presented by other alternative applications.
3 METHODOLOGY
The proposed system consists of a Multi Objective
Genetic Algorithm coupled with a market return
evaluation module that does the fitness evaluation,
and this, based on the estimation of the two
conflicting objectives, on the chosen market, and on
the specified period.
3.1 Strategy and Parameters
The strategy tested on this work was the Moving
Average Crossover (MAC), which is based on the
use of two Moving Averages (MA), with different
periods. One, formed by the MA with the shorter of
the two periods is called the "Fast MA”, and the
other, with the longer period is the "Slow MA". The
"Fast MA" reflects changes earlier than does the
"Slow MA". A buying (or sell short) signal is
generated when the Fast MA crosses over the Slow
MA. Conversely, sell (or a buy short) signal is
generated when the Fast MA crosses under the Slow
MA.
After defining the strategy, it is necessary to
define the parameters of the MAC, which in the case
are the type of the MA’s and the corresponding
period. It is important to stress that, for the type of
MA to use, the GA has also the freedom to choose
between a Simple or an Exponential MA.
Although it is common to tune the parameters of
one single TI and then use it to generate buy and sell
signals, for both long and short positions, in this
article, the option of using a separate set of
parameters for each of the possible actions was
taken; to specify: "enter long"; "exit long"; "enter
short"; and "exit short".
Some pre-processing of the historical data is also
done. This applies for instance to the MA periods,
which are calculated at program start and are limited
to the following set of Simple or Exponential MA's:
1, 4, 8, 12, 14, 16, 20, 24, 28, 32, 36, 40, 55, 60, 65,
70, 75, 80, 85, 90, 95, 100, 110, 120, 130, 140, 150,
160, 170, 180, 190, 200 and 250 days. This set of
periods has been chosen because it covers the most
widely used, long and short-term MA periods, found
Multi-objectiveOptimizationofInvestmentStrategies-BasedonEvolutionaryComputationTechniques,inVolatile
Environments
481