Mining Research, Based On The Uncertainty Of The U-ahc Co-location Model

Uncertain data is inherent in some important application fields, such as sensor networks and mobile object tracking. With the rapid development in data gathering and processing, the uncertainty of data has been studied gradually in-depth. Data mining based on uncertain data and the knowledge discovery caused many scholars interest, this more in line with the objective reality theory method has become more and more research hotspot in recent years. However, using traditional mining methods on uncertain data will bias the answer set, even is wrong, and hence cannot satisfy users'needs. So it becomes more and more important that how to effectively analyze the large amount of uncertain data in uncertain databases to discover the potential, valuable, and interesting information.The spatial co-location pattern mining discovers the association relationships among the spatial events, which is an important branch in the field of the spatial data mining. Different from mining on certain data, based on uncertain data co-location mining will introduce probability density function to describe the uncertain object position uncertainty, and measure two uncertain objects'distance by expectation distance(ED).In this paper, we analysis the mining methods with uncertain data firstly, and introduce the research results about classification, clustering, association rule and outlier detection. Then, explain the concept of co-location pattern, describe the process of co-location pattern mining in detail, also discuss the class algorithms of co-location mining, such as join-based and order-clique-based. Secondly, we summary the mining methods of co-location pattern, and propose a algorithm of mining co-location pattern based on Agglomerative Hierarchical Clustering(AHC), which pruning some instances that the distance between them greater than distance threshold by computing the clusters satisfy the coarse neighborhood relationship, this work finish based on the result of clustering, and the experiment shows that it is correct and efficient. Finally, this algorithmic is extent to uncertain data, i.e. the U-AHC is presented, and we simplify the two uncertain objects'distance computation via expectation squared distance, which saves time obviously, the final step is mining co-location pattern based the clustering result with uncertain data. At last, the conclusion and future work were presented.