On Futuristic Query Processing in Data Streams

Recent advances in hardware technology have resulted in the ability to collect and process large amounts of data. In many cases, the collection of the data is a continuous process over time. Such continuous collections of data are referred to as data streams. One of the interesting problems in data stream mining is that of predictive query processing. This is useful for a variety of data mining applications which require us to estimate the future behavior of the data stream. In this paper, we will discuss the problem from the point of view of predictive summarization. In predictive summarization, we would like to store statistical characteristics of the data stream which are useful for estimation of queries representing the behavior of the stream in the future. The example utilized for this paper is the case of selectivity estimation of range queries. For this purpose, we propose a technique which utilizes a local predictive approach in conjunction with a careful choice of storing and summarizing particular statistical characteristics of the data. We use this summarization technique to estimate the future selectivity of range queries, though the results can be utilized to estimate a variety of futuristic queries. We test the results on a variety of data sets and illustrate the effectiveness of the approach.