Research On Formal Description And Modeling Of Chance Discovery

CD (Chance Discovery) and CM (Chance Management) is a new research area originated since about 2000.Because of the novelty of CD&CM, many questions have not been properly answered, such as: Features of Chance, The necessity for CD&CM to be an independent research area, The relationship between CD&CM and the other traditional disciplines, including Artificial Intelligence, Economics, Decision Theory, Cybernetics System Theory and KDD etc. The techniques adopted to characterize CD&CM, Some drawbacks in these techniques, All these questions need further discussions.At first, we present some tentative answers for the above questions and indicate that the main goal of this thesis is to characterize the basic speciality of Chance by some mathematical techniques. In the following, we adopt different tools to characterize the speciality of Chance and the inference of Chance Discovery from three different aspects.We introduce the philosophical background, formal model, hypotheses generation and selection of Abductive Reasoning to explain the hypothetic features of Chance with Abductive Reasoning. On the basis of which we introduce how to explain Chance in the framework of Abductive Reasoning.Then we describe the relevance of Chance in the family of L_m4c. We firstly analyze the principle of CD-Relevance: if an event or a situation is the Chance of a goal of an agent, then it must satisfy the condition of CD-Relevance. On this basis, we propose a four-value formal system, L_m4c and its extensions. On the other hand we define the semantics of CD-Relevance directly and prove the equivalence of these two representations. Finally, we discuss the integration of L_m4c. with the Abductive Reasoning and then achieve some new good features.At last, this paper pays attention to elaborate KeyGraph algorithm and its characteristics as a tool of chance discovery. After analysis, an approach of KeyGraph parameters optimization based on small-world theory and genetic algorithm is proposed, in order to achieve automatic tuning of the KeyGraph parameters and avoid users focus and fatigue. At the same time, a mapping algorithm to extract the data referred to in the scenario and annotate those is proposed, for further data analysis as well as for supporting group discussion and resolving the ambiguity involved in user's interpretation of the graph, the results are important in chance discovery process. Furthermore, we introduce KeyGraph to double helix model, in which humans and data-mining tools co-work, to convince us double helical model of chance discovery, on the basis of it, a data-mining framework for chance discovery to connote chance discovering processs on the double helix model is proposed. This paper also makes improvement for this model from the viewpoint of decision-making, and proposes an IGA with on-chance operator to recognize a chance path for satisficing solutions.