Crane Monitoring And Data Mining System (qd-minner)

With the rapid development of information technology, more and more enterprises have relied on computer and storage. The ability of accumulating data has been up to amazing stage. So business and science research field are eager to find an abstracting scheme to find the rule and the technology of dependence between data. Now bank and retails industries have begun to utilize these sophisticated technologies to gain important knowledge. Moreover, to integrate OLAP, DSS, data mining and alarm system forms a nowadays very popular software- Bl (Business Intelligence).Under the background of the project data mining for crane inspecting information and forecasting technology, the paper have developed an specializing datamining system QD-Minner according to data mining technology. The mining object is special compared to others because they are the association rule between each inspecting node which are reflected as the association rule between columns in the database.Main mining approaches of QD-Minner are based on K-Means and FP-Tree. The paper have do some improvements to correct some limitation of these two kind approach. The improvements as below:(1) In the practical utilization of K—Means, the cluster number K is not a constant which changes each time. Because people can get proper partitions once, while changing k on and on for gaining the best partition. According to the analysis to K—Means, we can utilize the similarity of last time partition to calculate the next time partition. This is BLK—Means. The algorithm is based on the last partition to avoid the blindness of random choosing points. The advantage of this algorithm can be reflected as reduction of iteration times & runtime.(2) Though FP-Tree has gained a big improvement in the aspect of candidate items compared to Apriori, it still can be improve in the aspect of reading speed. The paper utilizes the advantage of bit-map, comes out with BMFP algorithm. The improvement makes the whole process can only read data once from database.