Mining high-dimensional data, sequences and data streams / Prospection de données de grande dimension, de séquences et de flux de donées

近年来，数据库的规模迅速增加，能够持续产生大量数据流的数据源变得越来越普遍。例如，雅虎主页每天收到1.66亿次点击，每小时约有48 Gb的点击流收集。其他例子包括交通和天气数据、卫星观测、银行交易和通过高通量技术进行的基因组测序。越来越需要有效的数据挖掘方法和知识发现系统来利用这些数据。聚类分析是数据挖掘的一个重要分支，其目的是将数据分成有意义的组。对高维数据、序列数据和数据流进行聚类是非常困难的，因为数据对象之间的距离/相似性的常规度量是没有意义的，这些数据集中的结构是复杂的，并且通常需要真实的时间处理。如果数据类型是分类的而不是数字的，这些困难会进一步扩大。

In recent years, the size of databases has increased rapidly and data sources capable of continuously producing large flows of data are becoming increasingly common. For instance, the Yahoo home page receives 166 million hits per day and about 48 Gb of clickstream collection per hour. Other examples include traffic and weather data, satellite observations, banking transactions and genomic sequences via high-throughput technologies. There is an increasing need for efficient data mining methods and knowledge discovery systems to utilize this data. Cluster analysis, which is aimed at separating data into meaningful groups, is a major sub-domain of data mining. Clustering high-dimensional data, sequence data and data streams is very difficult because conventional measures of the distance/similarity between data objects are not meaningful, the structure in these data sets is complex and real time processing is often required. These difficulties are further amplified if the data type is categorical rather than numerical.