Making Profit in Stock Investment Before XD Dates by Using Genetic Algorithm - Innovative Management in Information and Production - page 330

When some companies announce unexpected dividend rates, especially the
higher rates than the expected ones, then their stock prices will shoot sharply
before the XD dates. Therefore, investors who collect the stocks before the rise
of their prices will make substantial profit. From observations of many stocks, this
phenomenon is possible to happen. If an investor knows how many days before XD
dates are on the lower regime, then he will make good profit in the speculation of
the stock price by purchasing the securities on those days.
This study will use genetic algorithm to detect the regime switching and identify
the lower and higher regimes of eight stock prices before XD dates in the stock
exchange of Thailand. It will suggest buying strategies for those stocks. It will also
evaluate the performance of the suggested buying strategies. The research results
will empower investors to buy stocks at the right time and make them successful in
speculation in the stock market.

2 Development of the Model

There are several reasons why this study chooses genetic algorithm over other
econometric models. First, GA is a model-free method. It does not rely on the
type of data distribution and the stationary of the data. Therefore, it is flexible to
use for all time series. Second, GA ensures profit maximization. It is a tool for
optimization. It finds a solution that maximizes the objective function. It can ensure
that the regime switching is from a lower to higher regime. It avoids the solution
which indicates the switching from higher to lower regime. Third, GA is good to
apply to a specific part of time series. It does not need a longtime series data: 30-50
observations are enough. Last, GA does not aim at forecasting. Rather, it emphasizes
on what should be done during the whole process to ensure the highest performance
or output. Therefore, it is good for indicating when the appropriate time periods to
buy the stock are.
This study extends the work of Sudtasan ( 2012 ) which is the pioneer in applying
genetic algorithm to detect regime switching in stock prices before “window
dressing” at the year end. Before the work of Jann ( 2000 ) used genetic algorithm to
detect multiple change point. This is the first paper that identify clearly that he uses
genetic algorithm for the detection of the change point of a time series. Li and Lund
( 2012 ) also used genetic algorithm to detect multiple change point of a time series
of climate data.
Considering previous studies on detection of regime switching and change points
using variety of quantitative methods, this study proposes to use genetic algorithm
to directly detect the regime switching. It is quite similar to Sudtasan and Suriya
( 2012 ), but it improves many features such as a better crossing-over process and
clearer criteria to detect the regime switching. It differs from Jann ( 2000 ); Li and
Lund ( 2012 ) in that it uses a shorter time series to detect the regime switching in
a particular part of the series. Moreover, it focuses on detection of a single change
point rather than multiple change points.