Research On Mining Bidirectional Association Rules Of Temporal Dataset

Data mining researches the process of extracting implicit, previously unknown, and potentially useful information from large quantities of data. With the explosion of the current data, large datasets in electronic form than at any time heretofore. Various techniques have been employed to convert the data into information, including clustering, classification, regression, association rule induction, sequencing discovery, and so forth. We pay more attention to the temporal data to avoid processing huge data in real database. New knowledge discovery from temporal databases is a popular task in data mining.Temporal data mining has become an important branch and quite study circumstance of data mining field. At present, most of the studies about its association rule mining lack united theoretical frameworks, and model and corresponding algorithm based on temporal data are only applied to some special date type. What's more, along with introducing of fuzzy set and rough set theory, how to applied uncertainty theory to mine isn't also solved etc. In this thesis, at first, by means of analysis and study, Association rules mining is identified as one of the important problems in data mining. Let us first define the problem for a database containing a set of transactions, where each transaction contains a set of items. An association rule has the form of where and are sets of items. In such a rule, we require that the frequency of the set of items is above a certain threshold called the minsup. The frequency (also known as support) of a set of items is the number of occurrences of in. or the number of transactions in that contain. The confidence of the rule should also be above a threshold. By confidence we mean the probability of given.Time is also an important factor in real databases. We should extract useful information under the strict application of temporal database. So we also present a temporal relation model restricted within time limits. The enterprise need to analyze the amount of data to form an exact estimation of customs' purchase power during a particular period of time. Based on the positive association rule and the notion of the time-model, we propose a new kind of one called bidirectional association rule. It denotes a strong association among data in the temporal database.