Transactions on Computational Collective Intelligence XII

Until now, the use of attribute tables, which enable approximate reasoning in tasks such as knowledge integration, has been posing some difficulties resulting from the difficult process of constructing such tables. Using for this purpose the data comprised in relational databases should significantly speed up the process of creating the attribute arrays and enable getting involved in this process the individual users who are not knowledge engineers. This article illustrates how attribute tables can be generated from the relational databases, to enable the use of approximate reasoning in decision-making process. This solution allows transferring the burden of the knowledge integration task to the level of databases, thus providing convenient instrumentation and the possibility of using the knowledge sources already existing in the industry. Practical aspects of this solution have been studied on the background of the technological knowledge of metalcasting.

The paper focuses on the automatic opinion analysis related to web discussions. First, the paper introduces an approach to the extraction of texts from web forum discussion contributions and their commentaries as well as filtering these texts from irrelevant and junk information. Second, it describes variety of approaches to the sentiment analysis problem. It introduces a method for solving basic problems of opinion analysis (determination of word subjectivity, polarity as well as intensity of this polarity). The method solves the reversion of polarity by negation as well as determination of polarity intensity of word combinations. A dynamic coefficient for the word combinations processing is introduced and an implementation of the method is presented. In addition, the paper describes test results of the presented implementation and discussion of these results as well.

Finding the adequate negotiation strategy with incomplete information, even in one to one negotiation, is a complex problem. Inspired from research works aiming to analyze human behavior and those on social negotiation psychology, the integration of personality aspects, with the essential time parameter, is becoming necessary. For this purpose, first, one to one bargaining process, in which a buyer agent and a seller agent negotiate over single issue (price), is developed, where the basic behaviors based on time and personality aspects (conciliatory, neutral, and aggressive) are suggested. Second, a cognitive approach, based on the five-factor model in personality, is suggested to control the resulting time-personality behaviors with incomplete information. In fact, the five factors are the extraversion, the agreeableness, the conscientiousness, the neuroticism, and the openness to experience. Afterwards, experimental environments and measures, allowing a set of experiments are detailed. Results, concerning time-personality behaviors, demonstrate that more increasing conciliatory aspects lead to increased agreement point (price) and decreased agreement time, and more increasing aggressive aspects lead to decreased agreement point and increased agreement time. Finally, from a study case, of three different personalities corresponding to three different cognitive orientations, experimental results illustrate the promising way of the suggested cognitive approach in the control of the time-personality behaviors.

The paper deals with the problem of offering predictive service in e-commerce Web server systems under overload. Due to unpredictability of Web accesses, such systems often fail to effectively handle peak traffic, which results in long delays and incomplete transactions. As a consequence, online retailers miss an opportunity to attract new customers, retain the loyalty of regular customers, and increase profits. We propose a method for priority-based admission control and scheduling of requests at the Web server system in order to differentiate Quality of Service (QoS) with regard to user-perceived delays, i.e., Web page response times provided by the system (as opposed to HTTP request response times). To detect and cope with the system overload, a new kind of a load indicator is proposed, based on online measurements of page response times. Simulation results demonstrate that our solution is capable of providing key customers with limited delays while improving QoS for ordinary customers under heavy load.

Patent analysis has been recognized as an important task at the government and company levels. Patent data contain plentiful technical information, which is worthwhile to be used in patent analysis in order to find out the technical categories and the technological trend. Due to the complex nature of patent data, two data mining methods: IPC-based clustering and link analysis, are used to figure out the possible technological trend on thin-film solar cell. IPC-based clustering, a proposed clustering method for exploiting the professional knowledge of the patent office examiners, will be utilized to generate the IPC-based clusters via the IPC and Abstract fields; while the link analysis will be adopted to draw a link diagram via the Abstract, Issue Date, and Assignee Country fields. During experiment, the major technical categories will be identified using IPC-based clustering, and the technological trend will be recognized through the link diagram. Finally, the major technical categories and technological trend will be provided to the managers and stakeholders for assisting their decision making.

LCC is a Lightweight Coordination Calculus which can be used to provide an executable, declarative specification of an agent interaction model. In this paper, we describe an LCC-based system for specifying the migration behaviour of virtual machines within, and between datacentres. We present some example models, showing how they can be used to implement different policies for the machine allocation and migration. We then show how LCC models can be used to manage the workflows that involve creation and deletion of virtual machines when migrating services between different datacentres.

Place/transition Petri Nets (P/T PN) are utilized here to: (i) model modules of the endangered area (EA) having some attributes of simple agents; (ii) coerce the autonomous modules to a cooperation by means of supervision; (iii) model workflow of the evacuation process from EA. The approach is applied to analyzing the possibilities of evacuation EA being a part of a building and to finding the suitable free and safety escape routes. A supervisor is synthesized in order to force modules (agents) into a cooperation. The supervisor represents a desired goal of the cooperation and ensures its achievement. Thus, it makes the cooperation of modules possible. Its synthesis is based on the supervision methods known in DES (discrete-event systems) control theory. A simple case study of the building evacuation illustrates the proposed approach. The P/T PN based modelling the EA and the P/T PN-based evacuation workflow are proposed there. The differences between them as well as the possibilities of their mutual complementing are pointed out.

This paper introduces a modified particle swarm optimization (PSO) that exhibits the so-called “extreme social disagreements” among its wandering particles in order to resolve the stagnation when it occurs during search. We provide a short theoretical introduction about particle swarm optimization, then we describe and test our modified algorithms. We conclude from tests on several optimization benchmarks that our approach may help PSO escape stagnation in most of the situations in which it was tested. This work is intended to illustrate one of the benefits of using disagreements in social algorithms like PSO.