Jayantha Rajapakse

Fast growing self organizing map for text clustering

This paper presents an integration of a novel document vector representation technique and a novel Growing Self Organizing Process. In this new approach, documents are represented as a low dimensional vector, which is composed of the indices and weights derived from the keywords of the document. An index based similarity calculation method is employed on this low dimensional feature space and the growing self organizing process is modified to comply with the new feature representation model. The initial experiments show that this novel integration outperforms the state-of-the-art Self Organizing Map based techniques of text clustering in terms of its efficiency while preserving the same accuracy level.

Conceptual modeling in practice: An evidence-based process-oriented theory

Interviews were undertaken with 26 highly experienced conceptual modelers to determine how they undertook conceptual modeling in practice. Major types of cognitive states and events experienced by the conceptual modelers, as manifested in typed transcripts of the interviews, were then identified and coded. Based on the codes and the interview transcripts, a descriptive, process-oriented theory of how the modelers undertook conceptual modeling was induced.

e-Government implementations in developing countries: Success and failure, two case studies

e-Government is now used as the key instrument to provide e-services to the citizens and encourage the citizens to engage with the government on policy decisions through on-line inputs. Even though e-Government adoptions are wide-spread results have been mixed. In this paper, one successful e-Government initiative and one failed initiative in two developing countries are presented. During April 2007 the South African government launched the electronic National Traffic Information System (E-Natis). In this paper the researchers documented their investigative process and findings regarding the E-Natis debacle. This negative case study was compared with a similar case study in Sri Lanka in the same type of operation where the implementation of e-Revenue license initiative was successful.

Understanding individual play sequences using growing self organizing maps

When interacting with computer games, a player would derive a strategy to conquer the game using his/her experience, observations etc. During game play, the players adapt their strategy to better suit the challenges posed by the game. With time, these strategies would formulate to a pattern of interaction for an individual player with respect to myriad of game entities such as on handling of artificial opponents, movement strategies and decision making. Understanding these patterns would provide valuable insight about a players approach toward defeating the game which could be exploited to enhance the level of challenge posed by game AI. This paper attempts to identify dominant game play sequences made by an individual player by interpreting the positioning of the clusters in a growing self-organizing map (GSOM) generated using play data collected from the same player. Results indicate that dominant play sequences could indeed be identified but requires further analysis before a solid claim in this regard could be made.

Logical database design with ontologically clear entity relationship models

It is argued that ontologically clear entity relationship models can model the real world domains more accurately than ontologically unclear models. However, transformation of such models into the relational model at the logical level has not yet been studied sufficiently with a view to formulate new transformation rules. This paper presents a set of new transformation rules to convert ontologically clear conceptual models to relational models. Finally we did a comparison of two relational models that were developed from the ontologically clear and unclear models using a quality criterion synthesized from the extant literature. The preliminary results of this ongoing research study shows that the quality of relational model developed from ontologically clear conceptual model is superior to its ontologically unclear counterpart.

Can ERP Adoptions Change Organisational Culture in Developing Countries in Asia? An Empirical Investigation

Little research has been conducted to investigate a possible transformation of organizational culture through western-developed Enterprise Resource Planning (ERP) systems in developing countries in Asia. A case study carried out in Sri Lanka found that ERP adoptions along with other appropriate mechanisms can be a catalyst for the transformation of organizational culture. A little over 10 years after going live, the organizational culture that has inhibited the achievement of benefits from the western-developed ERP system is slowly changing at the case organization. Cultural transformation (user-behavior pattern change) in the case company occurred by accepting three out of four user-behavior patterns assumed by the western-developed ERP systems to a noticeable extent. Furthermore, case organization used several mechanisms such as company-wide education programs, phased-out implementation approach, recruiting an increasing number of young professionals to facilitate the transformation process taking place through the ERP adoption. Additionally, from this evidence, it can be deduced that organizational culture may not be static, over time they may be changed through various mechanisms.

Brain-inspired self-organizing model for incremental learning

Machine learning techniques which are involved in knowledge extraction from stationary datasets have been becoming inefficient due to the dynamic nature of contemporary data spaces. Hence, machine learning research constantly investigates incremental learning techniques to address this requirement. However, it is always a challenge to uncover useful information incrementally from a non-stationary input space because of the complexity an algorithm introduces to counter the stability-plasticity dilemma. In order to facilitate this demand a learning model is proposed using the self-organization and competitive learning strategy. Moreover, an algorithm which is implemented based on the proposed model is also presented with the experimental results to prove the validity of the proposed learning model in a non-stationary context.

Investigating individual decision making patterns in games using growing self organizing maps

Each player has an unique and a distinct way of interacting with a computer game due to the preconceived notion and the experience gained through playing the game title. During the game play a player adapts to the challenges posed by the game and a pattern of interaction emerge corresponding to factors such as tackling opponents, movement strategies and even decision making at certain game environments. Understanding decision making patterns provide valuable information about the players which could be exploited to enhance the total game play experience. This paper investigates the possibility of understanding the decision making patterns of a player whilst playing the 2D arcade game Pac-Man using an unsupervised approach known as the growing self organized map (GSOM). Results of this study motivated us to conjecture that player decision making patterns could be identified and explained via unsupervised learning.

e-Government adoption in developing countries: Lessons learned from a Sri Lankan case study

e-Government initiatives in developing countries are still in infancy and struggling to catch up with their counterparts in the developed world. In this paper findings from a case study in Sri Lanka, a developing country, are presented as a set of lessons learned. These lessons may shed some light into new e-government initiatives in developing countries with similar environments to Sri Lanka.

Evaluation of the Ontological Completeness and Clarity of Object-Oriented Conceptual Modelling Grammars

SeveralresearchstudieshaveconcludedthatmodellinggrammarsthatsupporttheObject-Oriented (OO)methodologyfocusmoreonmodellingsystemdesignandimplementationphenomenathan real-worldphenomenainISusers’domains.Thus,thepurposeofthisresearchstudywastoevaluate thesuitabilityofOOmodellinggrammarsforconceptualmodelling.Althoughtheresearchwork focusedononewidelyusedOOmodellinggrammar—namely, theUnifiedModellingLanguage (UML)—theapproachdevelopedcanbeapplied toanyOOmodellinggrammar.Thefirstphase ofthisresearchstudyfocusedonevaluatingallUMLconstructsandidentifyingasubsetofUML constructsthatarecapableofrepresentingreal-worldphenomenainuserdomains.Thesecondphase wasanempiricalevaluationoftheidentifiedsubsetofUMLconstructs.Theresultsofthisempirical evaluationsuggestthatinsteadofusingallUMLconstructsthesubsetofUMLconstructsisbetter suitedforconceptualmodelling. KEywORDS Bunge-Wand-Weber (BWW) Ontology, Canonical Action Research (CAR), Conceptual Modelling, Ontological Completeness and Clarity, OO Modelling Grammars, UML