Semi-supervised Clustering Using Degree And Its Application In The Short-term Forecasting Of Export-container Quantity

Data mining is to discovery the interesting information and knowledge, which is useful, connotative, and unpredictable, from numerous data. Presently, data mining has attracted lots of attentions and becomes one of the international advanced directions in the field of database and information decision.As a frequently used technology in data mining, clustering's object is to sort cases into groups, or clusters, so that the degree of association is strong between members of the same cluster and weak between members of different clusters. Clustering is unsupervised classification analysis which won't benefit from the assistance of the teacher. However, with the rapid development of the clustering research, people pay much attention to background knowledge. How to cluster integrated with users' knowledge has become one of the main challenges.First of all, the paper analyzes the probabilistic model for semi-supervised clustering and related algorithm. The algorithm ignores the cannot-link constraints and has poor performance. A semi-supervised clustering using degree(SCUD) algorithm is proposed. The main idea of method is initialize the initial cluster center based on constraint degree, which is calculated by the given background knowledge in the form of must-link and cannot-link constraints and then cluster with EM algorithm. A series of experiments show that the SCUD algorithm have better cluster performance in the condition of relatively fewer constraints, and also show that it has lower time complexity when it is used in mining large databases.And then we propose a two-step forecasting framework which is the solution to the short-term forecasting of export-container quantity in the port. The forecasting algorithm using SCUD is presented based on the framework. The algorithm utilize the background knowledge, such as next port, maximum load of vessel, link constraints, to improve the forecasting precision. The precise forecasting will help the port administrator plan the container stack, the schedule of tyre crane and crane, personnel assignment in advance. And it will also greatly improve the efficiency of container operation, reduce general expenditure in the port.