C. O. Uwadia

Improving medical rule-based expert systems comprehensibility: fuzzy association rule mining approach

In this paper, a Fuzzy Association Rule Mining (FARM) with expert-driven approach is proposed to acquire a knowledge-base, which corresponds more intuitively to human perception with a high comprehensibility. This approach reduces the number of rules in the knowledge-base when compared with the Standard Rule-base Formulation (SRF) and makes possible the rating of the rules according to their relevance. The rule relevance is determined by the measures of significance and certainty factors. The approach is validated using a medical database and the result shows that this approach ultimately reduces the number of rules and enhances the comprehensibility of the expert system.

Improving Customer Relationship Management through Integrated Mining of Heterogeneous Data

The volume of information available on the Internet and corporate intranets continues to increase along with the corresponding increase in the data (structured and unstructured) stored by many organizations. In customer relationship management, information is the raw material for decision making. For this to be effective, there is need to discover knowledge from the seamless integration of structured and unstructured data for completeness and comprehensiveness which is the main focus of this paper. In the integration process, the structured component is selected based on the resulting keywords from the unstructured text preprocessing process, and association rules is generated based on the modified GARW (Generating Association Rules Based on Weighting Scheme) Algorithm. The main contribution of this technique is that the unstructured component of the integration is based on Information retrieval technique which is based on content similarity of XML (Extensible Markup Language) document. This similarity is based on the combination of syntactic and semantic relevance. Experiments carried out revealed that the extracted association rules contain important features which form a worthy platform for making effective decisions as regards customer relationship management. The performance of the integration approach is also compared with a similar approach which uses just syntactic relevance in its information extraction process to reveal a significant reduction in the large itemsets and execution time. This leads to reduction in rules generated to more interesting ones due to the semantic clustering of XML documents introduced into the improved integrated mining technique.

AMTEA: Tool for creating and exploiting annotations in the context of Economic Intelligence (competitive intelligence)

This paper presents annotation specification as attribute-value pair for representing and capturing strategic information for solving decision problems in the context of Economic Intelligence (also referred to as competitive intelligence). The aim of using this approach is to facilitate information reuse for similar problems. While most of available annotation tools allow annotator to add annotation to document of interest as atomic object, it is believed that annotation represented as an attribute-value pair will offer enhanced interpretation functionalities. Annotation specification integrating the identified needs of the annotator as well as partial implementation of our prototype called AMTEA (Annotation Model and Tools for Economic Actors) is presented. The system is designed for annotation creation and exploitation taking into consideration annotations made, decision problem being solved, and users profile.

Scheduling Cloud Workloads Using Carry-On Weighted Round Robin

Cloud Computing represents a paradigm shift in computing. It advocates the use of computing resources as a service rather than as a product. The numerous advantages which the Cloud offers has led to many users adopting it at a phenomenal rate. Providing service to this ever growing number of users in a fast and effective manner is a major challenge. Numerous researchers have proposed various approaches to scheduling user workloads, notable among which are the First-Come-First-Serve and Weight Round Robin (WRR), and have obtained varied levels of successes. Unfairness and excess allocation delay are some of the shortcomings of these approach. There is also the assumption that all Cloud users’ workloads belong to a single class of requirement. This work proposes an efficient and fair Cloud workload scheduling algorithm called Adaptive Carry-On Weighted Round Robin (ACWRR), and also takes into consideration multiple workloads classes. Experimental simulations were conducted with ACWRR benchmarked against WRR. Results show that ACWRR performs better than WRR by at least 13% in terms of system latency and 38% for makespan.