Stream Data Association Rules Mining Based On The Dynamic Support

There are some fields that generate stream data, such as industrial production, life activities, business transactions and so on. It is closely related to life, production. So it is very important to study the multiple data stream. Association rules mining under multiple data stream is a major aspect of stream data mining. Stream data association rules mining can find the relationship between different streams. Characteristics of the stream data include: massive, flowing and unpredictable, with the change of time, the data stream contains knowledge is constantly updated, so it is difficult to work on the stream data. Most of the time the people is more interested in the knowledge being contained in the lasted data, and care more about association rules occurred in different time segments, so this paper proposes inter transaction association rules mining method based on dynamic support threshold.At present, researchers have researched on the association rules mining for long, and the study of association rules is also divided into many aspects, and the different study methods can get different result, but most of mining methods of association rules get the intra transaction association rules in transaction. There is so little research on inter transaction association rules mining, and on the other hand there is so few methods to search a good support threshold.Firstly this paper uses the sliding window to limit stream data, then do preprocessing on stream data. In the processing of pretreatment using linearization method fitting to raw data and that can decrease the size of data set at the same time, and finally the end of the preconditioners is proposed in this paper to generate large transaction grouping method of inter transaction Asssociation rules. In this paper, in the process of mining association rules ITF-tree algorithm is proposed based on the FP-growth algorithm for mining association rules. With the increasing of time, the amount of data will be increased, but people will be lower interested than those of the recently generated data on the historical data of the attention in general, so this paper uses conceptual data attenuation, thereby reducing the influence of old data to the mining result. Due to artificial setting minimum support threshold may bring many problems, so this paper presents a method for searching minimum support threshold.Compared with existing algorithms on the thermal power plant data sets were compared test, we show that this method proposed in this paper is efficient. Can be very good to meet the requirement, Can be very well to meet the requirements of the stream data in time efficiency, also can mining inter transaction association rules effectively...