Knowledge Discovery Using Advanced Computational

							Journal of Information & Knowledge Management, Vol. 5, No. 4 (2006) 279
 c iKMS & World Scientific Publishing Co.




                                     Editorial on Special Topic Section

                           Knowledge Discovery Using Advanced
                             Computational Intelligence Tools


Data mining plays an important role in Knowledge Man-                (clustering). The developed algorithm iterates between
agement. Knowledge is the product of moving from                     clustering (assuming that the number of clusters is not
data to information and finally to knowledge. The Fifth               known a priori) and feature selection. Authors proposed
International Conference on Hybrid Intelligent Systems               two Bayesian approaches for feature selection (i) Na¨    ıve
(HIS’05) gathered individual researchers who see the need            Bayes Wrapper (NBW), and (ii) Markov Blanket Filter
for synergy between various intelligent techniques for vari-         (MBF). Experimental results reveal that NBW and MBF
ous data mining and knowledge management applications.               could reduce the number of features, while providing good
This special issue comprising of six papers is focussed              quality partitions.
on using advanced computational intelligence tools for                    In the fifth paper, Guajardo, Weber and Miranda
knowledge discovery. Papers were selected on the basis               propose a novel methodology for regression-based fore-
of fundamental ideas/concepts rather than the thorough-              casting. A generic hybrid approach is presented that iter-
ness of techniques deployed. The papers are organised as             atively selects the most relevant features and constructs
follows:                                                             the best regression model given certain criteria. The devel-
     In the first paper, Falkowski presents perceptron                oped model is suitable for feature selection as well as for
learning in the context of the so-called scoring systems             model construction. The application to several time series
used for assessing creditworthiness as stipulated in the             underlines its usefulness.
Basel II central banks capital accord of the G10-states. It               Nedjah and Mourelle, in the last paper, present a
is argued that the results obtained may be exploited to              novel deterministic multi-threaded complete matching
compute associated probabilities using a logistic activa-            method. Complete matching determines whether a sub-
tion function and maximum likelihood methods.                        ject term contains a sub-term that is an instance of a pat-
     Han and Cho, in the second paper propose a novel                tern in a pattern set. The developed method subsumes
method of predicting the user’s future movements in order            a deterministic lazy root-matching technique. The model
to develop advanced location-based services. The user’s              is evaluated using theorem-proving and DNA-computing
movement trajectory is modelled using a combination of               applications.
recurrent self-organising maps and the Markov model.                      The editors wish to thank the referees who have crit-
Future movement is predicted based on past movement                  ically evaluated the papers within the short stipulated
trajectories. A prototype application based on location              time. Finally, we hope the reader will share our joy and
prediction is also presented.                                        find this special issue very useful. We would like to take
     In the third paper, Abraham and Grosan propose                  this opportunity to thank Professor Suliman Hawamdeh,
an ensemble-based decision support system using genetic              Editor-in-chief, International Journal of Information and
programming (GP). The decision support system uses a                 Knowledge Management for all the timely advices and
combination of unsupervised learning for clustering the              help and also for providing an opportunity for editing this
data and an ensemble of three well-known GP techniques               important scientific work.
to classify the different decision regions accurately. Exper-
imental results reveal that the proposed ensemble method
performed better than the individual GP approaches and
the method is efficient.                                                                                       Nadia Nedjah,
     Hruschka et al., in the fourth paper, investigate the                             Ajith Abraham and Luiza M. Mourelle
adaptation of Bayesian methods for unsupervised learning                                                      October 2006




                                                               279