Database Reference
In-Depth Information
•
Step 4. Jurisdiction:
The court makes
decision and sentences based on relevant
laws. Most countries have made laws
for regulation in stock markets. For ex-
ample, Securities Disclosure of Interests
Ordinance (SDIO) is to regulate the insider
trading in Hong Kong. SDIO specifies the
definition of insiders and what constitutes
insider trading, and outlines the notifica-
tion procedure and the penalties for failing
to report (Wong et al. 2000).
is $5, and the price today is $6, then an alert will
be generated because the rise percentage is 20%,
larger than the predefined threshold 10%. For both
of the above cases, the prices rise by the same
amount, but they have different results.
The rule approaches have the advantage of
simplicity, and sometimes can effectively capture
some types of market abuse. However, the limita-
tion of the rule approaches is their low accuracy.
They tend to generate too many alerts and most
of the alerts are not related to any market abuse
or illegal behaviors.
technologIeS for excePtIon
MInIng In Stock MArket
Basic Statistic Methods
The technologies on stock market surveillance
have been developed for decades from the simple
rule-based technology to advanced artificial intel-
ligent technologies. Many researchers have made
contributions on stock market surveillance. The
well-known technologies for market surveillance
will be surveyed in this section.
By considering the movement trend of price or
volume over time, the basic statistical technique,
such as the mean and standard deviation is also
utilized in surveillance process (Smith 1995).
The use of the mean and standard deviation
assumes that the values for each measure are nor-
mally distributed. In order to reduce the number of
alerts generated, the standard deviations are tuned
accordingly. The traditional approach regards the
returns out of µ±3σ as outliers, where µ and σ are
respectively the mean and variance of Gaussian
distribution (Smith 1981). The stock measures,
however, are generally non-normal, which show
themselves as excess kurtosis and skewness.
The statistical tests are used to test whether the
price and volume data are normally distributed.
The stability of the distributions in different time
periods is also tested by statistical tests. However,
in practice, the price and volume movements of
most of stocks are not normally distributed. An
alternative way is to identify the distribution by
curve fitting technologies (Alhanaty & Bercovier
1998). In addition, the price and volume of most
stocks are not stable and it is difficult to identify
the underlying distribution over time. Therefore,
the basic statistic methods are inappropriate to
monitor stock movements.
rule Approaches
The rule approaches are the initial way to monitor-
ing stock markets. Normally a set of parameters
are predefined based on experience (Smith 1995).
If the real value is larger than the predefined pa-
rameter, an alert will be generated. In practice,
the price and volume are used to measure the
performance of stock. For example, assume that
the parameter of price movement is set to $2. If
the price of one stock is $8 on the first day, and
rises to $11 on the next day, an alert will be gen-
erated, because the increase is $3, larger than the
predefined threshold $2.
In order to improve the effectiveness of the
rule approaches, the threshold is often set as a
specific percentage to the prior value. For example,
assume we set the movement percentage param-
eter as 10%. If the price on the first day is $20,
and price on the next day is $21, then there is no
alert generated. However, if the price yesterday
Search WWH ::
Custom Search