Study And Implementation On Frequent Closed Cube Mining Algorithm Of Three Dimensional Microarray Data Sets

The human genome project (HGP) is put forward in order to have a more profound understand of human itself. The technique of biological experiment has been improved a lot with favoring development of HGP, in which microarray-chip technique plays an important role. Nowadays, microarray-chip technique can test the expressions of many genes of many samples during a series of time easily, and generates a three dimension microarray dataset.The improvement of experimental technique motivates the development of the technique mining informations from experimental results. Consequently, the technique mining frequent closed cubes is proposed. This thesis first proposes a novel algorithm mining frequent closed cubes, named MFCC algorithm, using the method of reducing dimension. MFCC algorithm first cuts the three dimension dataset into several two dimension datasets, and deals with them using certain frequent-pattern-mining algorithm; then intersects the results of the two dimension datsets to get all the frequent closed cubes with some pruning rules. The advantage of MFCC algorithm is that the efficient dimension-decreasing technique can not only decrease the dimension of the dataset, and use the existing algorithms mining frequent patters, but also avoid generating too many two dimension datasets, which bring a lot of two dimension mining tasks. In order to enhance the performance of MFCC algorithm, this thesis proposes an improved algorithm, named MFCC+algorithm. MFCC+algorithm uses the same dimension-reducing technique as MFCC algorithm, maintaining the advantages of MFCC algorithm. Some more efficient pruning rules are used in MFCC+algorithm, which can prun off all the unclosed cubes, avoiding checking the closeness of the mining results. The experimental results show that MFCC+algorithm has made a great improvement of MFCC algorithm, and the improved algorithm performs better than the previous algorithms.This thesis first introduces the background knowledge related to the research and the previous research works, then proposes the MFCC algorithm and MFCC+algorithm successively, and gives the correctness of the algorithms. The performances of the algorithms are tested through experiments at last.