Rule Strategies for Intelligent Context-Aware Systems The Application of Conditional Relationships in Decision-Support Philip Moore School of Computing Birmingham City University Birmingham, UK ptmbcu@gmail.com Bin Hu School of Computing Birmingham City University Birmingham, UK bin.hu@bcu.ac.uk Mike Jackson Business School Birmingham City University Birmingham, UK mike.jackson@bcu.ac.uk Abstract—Rule-based systems have been shown to enable automated problem solving and decision-support. Rule-based systems are however inherently domain specific requiring strategic rules design to address the requirements specification for the problem under consideration. Rule-based systems fall into two (general) types: traditional and fuzzy rule-based systems. This paper considers rule types and strategies with an introduction to logic systems as they apply to the approach proposed in this paper and an overview of related research. Consideration is given to the principles that underlie rule- based system design. The novel Context Processing Rules and Fuzzy Event Condition Action rules are presented with an evaluation and proof-of-concept. This paper posits that fuzzy rule-based systems provide an effective approach to enable context processing with constraint satisfaction in pervasive intelligent context-aware systems with decision-support. Keywords-Rules; Intelligent Systems, Information Systems, Context, Decision Support Systems, Personalization I. INTRODUCTION The primary function of pervasive intelligent context- aware systems is to enable the identification and selection of suitably qualified individuals for personalized service provision (hereafter termed personalization). Personalization requires the matching of potential recipients (hereafter termed the ‘user’) of a resource with a resource being distributed; for the purpose of this discussion a resource can be a document, file, link, or a request for an interactive online collaboration (hereafter termed the ‘input’). Personalization must however accommodate constraint satisfaction and preference compliance [1][2], hereafter termed constraint satisfaction (CS). Realizing personalization with CS requires the effective processing of contextual information. Contextual information describes an individual’s current personal and environmental profile (or context) [1][2][3]. Context has gained traction with the development of pervasive computing [2], a context reflecting the totality of contextual information that combines to identify an individual and in a pervasive mobile computing scenario, the network infrastructure and that of available mobile devices often implemented in ad-hoc wireless networks [1]. This paper posits that intelligent context aware systems offer the potential to effectively implement intelligent context with decision support to enable personalization with CS [4]. The paper is structured as follows. An overview of the requirements specification [for intelligent context-aware systems] is set out. Rule types and strategies are discussed with an overview of related research. The design parameters for Rule Based Systems (RBS) and rule engines are considered with a discussion on their application in intelligent context-aware systems. The Context Processing Rules (CPR) [1] and Fuzzy Event Condition Action (FECA) rules [5] are considered and the FECA rules algorithm is presented with an evaluation and proof of concept. The paper concludes with a discussion and open research questions. Figure 1. Personalization and the context-matching problem. Shown is the partial matching problem with a decision boundary (threshold) used in the CPR and FECA rules algorithm [2][5]. Note: the proposed approach enables multiple thresholds in for example context-aware health monitoring systems [6] where multiple decisions (prognses) must be accommodated. II. REQUIREMENTS SPECIFICATION The issues of personalization, intelligent context-aware systems, context processing, and context-matching with partial matching, and CS have been discussed and explored in [1][2][3][4][5][6] however the requirements for an intelligent context-aware decision support system can be summarized as follows: 1. The primary function of an intelligent context-aware system is to arrive at a Boolean decision as to the suitability of a user for personalization. Such systems can be viewed as decision support systems. 2. Arriving at a Boolean decision requires the matching of the input with the corresponding context properties a1 b1 b2 c1 c2 a1 b2 c2 b1 c1 Input Context Output Context Context Match Context Problem Potential Solution Context Decision Boundary (Threshold)