Computational Intelligence Applied to Financial Price Prediction: A State of the Art Review
| Autor | Javier Sandoval |
| Cargo | Docente investigador. Facultad de Finanzas, Gobierno y Relaciones Internacionales. Universidad Externado de Colombia |
| Páginas | 145-178 |
Computational Intelligence
Applied to Financial Price
Prediction: A State of the
Art Review
Javier Sandoval Archila*
javier.sandoval@uexternado.edu.co
* Docente investigador. Facultad de Finanzas, Gobierno y Relaciones Internacionales. Univer-
sidad Externado de Colombia. Magister en matemática aplicada de la Universidad Nacional de
Colombia. Magister en Finanzas de London School of Economics. Candidato a PHD en sistemas
y computación de la Universidad Nacional de Colombia.
ODEON Nº 6
147
PP. 145-178 • N.º 6 / 2011
1. Financial Price Prediction: Theory Review
The currently accepted paradigm in Finance has dominated the study of human
interactions in financial markets since Samuelson’s seminal work, “Proof that Pro-
perly Anticipated Prices Fluctuate Randomly” in 1965. The emerging theoretical
framework was built over three concepts: agents’ rationality, market efficiency
and neglected or zero predictability. This paradigm has been called Neoclassical
Finance and it has focused on studying market interactions in equilibrium, a steady
state, a state that does not offer any new stimulus to change. Therefore, relevant
questions have been what agents’ actions, interactions and strategies should be in
order to be with line to the steady state.
The established paradigm has had to overcome an indeterminacy presented with
its lines of reasoning. In financial markets, (as well as any other economic system)
individuals behave collectively to create an outcome. However, the same outcome
also influences the behavior previously mentioned. Therefore, the financial interac-
tion problem is ill-posed. As pointed out by Arthur (2006), Neoclassicals proposed
the following questions in order to overcome this problem. What prices and quan-
tities of financial assets offered and bought are consistent with equilibrium? Mo-
reover, what strategies and actions will also help to perpetuate this moveless state?
These two questions have defined the first pillar of modern Finance studies;
agent’s rationality and homogeneity. On one side, rationality and homogeneity
helped financial economics to developed tractable and analytical models. Howe-
ver, Finance has paid a high cost. Models created under the Neoclassical domain
are not necessarily supported by realistic assumptions. Therefore, conclusions are
elegant but sometimes wrong. Rationality drives financial economics to the so-
called expression, rational decisions or rational expectations. When participants
of a financial market make decisions, they continuously do so in accordance with
available information. Furthermore, strong rationality assumes that agents always
use the information at hand in the most optimal way.
Due to rationality, agents’ decision making process is reduced to an optimization
problem, hoping that individuals have the capacity to make necessary calculations
in an acceptable time. Ill-posed problems have been tamed. From a domain of po-
tential individual expectation models, the one that would validate the whole model
if every individual uses it, has been chosen. The answer is not difficult to guess,
with expectations following previous rationality definition. So, if every agent is
implementing the same decision making process, Neoclassicals rapidly came to
the aggregation assumption which states that every individual in the market can
Para continuar leyendo
Solicita tu prueba