Study On FP-growth Algorithm In Pervasive Computing Environment

The development of pervasive computing, cloud computing, communications networks and other technologies promoted the coming of the big data era. Among them, the ubiquitous computing network is the main source of data for big data. The potential value hidden in big data has become another nascent research hotspot of information society. One way to get the potential value is to make use of data mining techniques to implement knowledge discovery on data accumulated. Frequent patterns excavation, that association rules mining, is one of the important applications of data mining technology. It aims to discover the potential valuable relationship among the items.The typical algorithms of association rules mining are Apriori algorithm and FP-growth algorithm. The latter only needs to scan the databases twice, and thus having a better advantage than the former. In this paper, we put FP-growth algorithm as the research focus. In the pervasive computing environment, the heterogeneity of big data and the limitation of terminal equipment make the original serial algorithm or parallel algorithm based on a simple cluster unable to meet the demand. Therefore, this paper put the research of FP-growth algorithm based on cloud computing as the research core, in terms of the research of FP-growth algorithm in pervasive environment. Because, the cloud computing based on the Internet, the technical characteristics of it determine the possibilities of its being the technical support and processing platform of data mining of big data. Load balance control and data security control become the keys of study because of the technical characteristics of the cloud computing.Based on the study and analysis, this paper proposed a new load balance control strategy and an association rules mining model with relatively high data security. The three aspects included in the main tasks are as follows: Firstly, taking the total length of prefix path and frequency of each item as load metrics, to improve the load balance control strategy of FP-growth algorithm, and conducting experimental verification. Secondly, taking the hardware properties such as processing speed, memory capacity and disk capacity into consideration, to improve the load balance control strategy of cloud cluster, and implementing the theoretical analysis and verification. In the end, taking the technical characteristics of the ubiquitous computing network and cloud computing into consideration, to propose an association rules mining model with relatively high data security, and conducting the theoretical analysis and verification.In the end, this paper used the data sets generated by the IBM Data Generator to implement experimental verification, and confirmed that the improved FP-growth algorithm has higher efficiency than the existing algorithms such as BPFP. Meanwhile, we summarized the work that have done, and expounded the challenges and research prospect of big data mining theoretically.