Qazi Mudassar Ilyas

An Overview of Bayesian Network Applications in Uncertain Domains

 Abstract—Uncertainty is a major barrier in knowledge discovery from complex problem domains. Knowledge discovery in such domains requires qualitative rather than quantitative analysis. Therefore, the quantitative measures can be used to represent uncertainty with the integration of various models. The Bayesian Network (BN) is a widely applied technique for characterization and analysis of uncertainty in real world domains. Thus, the real application of BN can be observed in a broad range of domains such as image processing, decision making, system reliability estimation and PPDM (Privacy Preserving in Data Mining) in association rule mining and medical domain analysis. BN techniques can be used in these domains for prediction and decision support. In this article, a discussion on general BN representation, draw inferences, learning and prediction is followed by applications of BN in some specific domains. Domain specific BN representation, inferences and learning process are also presented. Building upon the knowledge presented, some future research directions are also highlighted.

Improving Usability through Enhanced Visualization in Healthcare

Data visualization has gained significant importance since the demand for software applications providing abstract view and pattern identification has increased. To meet the demand, a number of data visualization techniques and best practices have been proposed in many disciplines. However, data visualization is not as advanced in healthcare software applications as compared to other scientific fields. The result is a limited support provided to the healthcare practitioners by software applications. Moreover, poor usability of healthcare applications is another major hurdle in adoption of electronic healthcare systems rather than traditional paper based methods. In this paper, we propose an Electronic Health Record (EHR) system for Obstetrics, integrating different data visualization techniques. The proposed system has been evaluated using standard usability evaluation methods. The results indicate that using appropriate data visualization techniques in healthcare systems results in enhanced usability. The results also prove that better data visualization helps in improved quality of care and achieves high assurance in healthcare information systems.

A Base Level Ontology for Disaster Management

A disaster may be caused by natural or man-caused hazard and all disasters require some mitigation measures. Current Disaster Management Systems (DMS) depend on manual data entry, analysis, and decision making. Semantic Web technologies are gaining popularity for developing intelligent and self-governed systems and can be used effectively in a DMS. Ontologies are an integral part of a semantic DMS and in this article we propose two-fold use of the ontologies in a semantic DMS. Firstly, ontologies can be used as background knowledge for effective discovery and selection of resources. Secondly, after population, they can be used for reasoning to support decision making process. We propose a base level disaster ontology that covers prominent aspects of disaster management such as intrinsic properties of disaster (disaster type, disaster date and location), losses caused, services required, service providers, and relief items. The proposed ontology is evaluated using data-driven approach. The results of conventional crawling are compared with ontology-driven crawling on documents most-relevant, semi-relevant and irrelevant to disasters which show that ontology-driven crawling is more effective for resource discovery and selection as compared to conventional crawling.

A central tendency-based privacy preserving model for sensitive XML association rules using Bayesian networks

The rationale of XML design is to transfer and store data at different levels. A key feature of these levels in an XML document is to identify its components for additional processing. XML components can expose sensitive information after application of data mining techniques over a shared database. Therefore, privacy preservation of sensitive information must be ensured prior to signify the outcome especially in sensitive XML Association Rules. Privacy issues in XML domain are not exceptionally addressed to determine a solution by the academia in a reliable and precise manner. In this paper, we have proposed a model for identifying sensitive items nodes to declare sensitive XML association rules and then to hide them. Bayesian networks-based central tendency measures are applied in declaration of sensitive XML association rules. K2 algorithm is used to generate Bayesian networks to ensure reliability and accuracy in preserving privacy of XML Association Rules. The proposed model is tested and compared using several case studies and large UCI machine learning datasets. The experimental results show improved accuracy and reliability of proposed model without any side effects such as new rules and lost rules. The proposed model uses the same minimum support threshold to find XML Association Rules from the original and transformed data sources. The significance of the proposed model is to minimize an incredible disclosure risk involved in XML association rule mining from external parties in a competitive business environment.

Learning Bayesian Network leveled-structure from support based XML frequent itemsets

XML (eXtensible Markup Language) is a standard and entirely user-driven language for storage and transfer of information. XML frequent itemsets are usually found for mining XML association rules from XML transactional databases. These XML frequent itemsets lead researchers to find interesting XML patterns in large databases with the use of a threshold value. Apriori algorithm is one of the most leading solutions to discover XML frequent itemsets based on support value. XML frequent itemsets consist of similar items which show evidence of association. This relationship can be found with the use of Bayesian Network by learning structure of XML frequent itemsets. K2 algorithm is used to learn the structure of XML frequent itemsets. In this work, we propose a novel Apriori K2 algorithm. This algorithm is composed novel direction of apriori and K2 algorithms to find XML frequent itemsets and learning a level-wise Bayesian Network structure. For learning each level of this structure, XML frequent itemsets are found from XML candidate itemsets with the use of support measure using apriori algorithm. An updated binary table is prepared based on XML frequent itemset during the execution of apriori algorithm. K2 algorithm is used in conjunction with apriori algorithm to learn Bayesian Network structure of XML large frequent itemsets and find their relationship at each level. We have extensively tested our solution over UCI machine learning datasets and measured its performance. The results have shown that performance of our proposed solution is better than the combined performance of apriori and K2 algorithms.