Research On Modeling And Mining Problems Of Indeterminacy Temporal Data Based On Probability

Temporal Database is adopted as an object of study in the paper. Meanwhile,research on the indeterminacy temporal data is focused. Many applications such as AI,database management, multimedia system, history management system, medicalinformatics, etc., inevitably encounter indeterminacy temporal data because of thedynamic changes of the real world. They can not be efficiently represented and stored indatabases, especially the valid time of some incidents and their temporal relations can notbe accurately determined. For solving above-mentioned problems, modeling, queryingand mining are studied.(1) BPTM (Temporal model based on probability) is proposed. In this paper, throughthe analysis of the existing indeterminacy temporal information model and usingprobabilistic method proposed by Snodgrass, BPTM is designed. In this model, someoriginal predicate functions are extended and redefined, supporting not only theexpression of indeterminacy temporal information, but also expression indeterminacyrelations between temporal information by using the probability.(2) To amend the fuzziness parameter in BPTM model. The value of fuzziness isgiven according to the experience in the field by the user of a real number between0to1.In fact, each user’s experience is different, so the query result will also have differences.In order to enable users to obtain information more universal and practical, first by theuser input fuzzy initial value, and then use the Bias classifier to modify the fuzzy degree,so that users get information more realistic and practical.(3) A Medical Data Mining Algorithm Based on Discrete Particle SwarmOptimization is proposed (PSO-BNB). Based on the existing algorithm of attributeselection process comparison, and according to the medical data is huge and thecharacterization is not obvious, a method of discrete particle swarm optimization isproposed to solve the attribute selection process for improving the classification accuracyof medical data. First of all, discretization of the medical data is finished; then, thewrapper method based on PSO is used to carry out attribute selection of discrete datasets, and Na ve Bayesian classifier is used to train the selected attribute subset. Theexperimental results show that the algorithm has higher classification accuracy.(4) Based on the BPTM, how to represent indeterminacy temporal information andhow to query and mine the indeterminacy relations between temporal information are solved. In terms of the requirement of storing and processing medical temporalinformation, the temporal DBMS technology and BPTM are applied to the medical field.The temporal information storage and its representation are solved by increasing thespecific temporal module in relational database. Meanwhile, querying and mining issolved. By using probability and PSO-BNB, we can mine the temporal information...