Abstract:
Generally, stock markets represent the economic status of
a country. Different stock markets have different
characteristics based on the economics of the countries
they represent [1]. Stock markets are complex, nonlinear,
dynamic and chaotic [2]. Prediction of stock prices is one
of the major challenges in the hands of business analysts
for the interest of the financial and economic personnel
and in general the policy makers and planners of a
country. Techniques of prediction vary greatly according
to the availability of information, quality of modeling and
the underlying assumptions used. Neural networks are
regarded as more suitable for stock prediction than other
techniques mainly due to its ability to adaptation to the
nature of observations (data); for instance in situation like
financial markets in general it is impossible to incorporate
the effects of some socio-economic factors directly in the
models (e.g. Exchange rate of other non-western
currencies, GDP, Consumer indices, etc.) [1]. We used the
S&P SL20 index of Colombo Stock Exchange (CSE) in
Sri Lanka and exchange rate between US $ and LKR data
from 02nd January 2013 to 31st December 2013. The best
feed forward ANN model and suitable parameters of each
model are selected by using training data set with trial
and error technique. The accuracy of each model was
compared via Absolute fraction of variance (R2), Mean
Absolute Deviation (MAD), Mean Square Error (MSE),
Root Mean Square Error (RMSE) and visually. The
results examined that the exchange rates which is an
external factor affected the share price forecasting in
ANNs.
