The Approximate Query Research Of Time Series Based On Linear Hash Index

The time series is a quite important data type among the temporal data,it produced continuously by applications in many fields.At the same time,it also contains a lot valuable information.It not only can provide decision support for managers,but also provide effective forecast for the production trends in the future.Therefore,how to deal with the time series is a hot issue both at home and abroad.However,because of the complexity and the large quantity of time series,the index creating and query processing are unable to based original time series directly.It' s necessary to preprocess the time series before we use it.In this paper,a normalized preprocess method has been put forward to realize the discrete presentation of the time series.Aimed at the problem that approximate query to the time series has a high time com-plexity.In this paper,we propose a time series indexing technique based on linear hashing.What we used during the processing of creating index are the discrete presented time series.The discrete presentation doesn' t a single representation of a time series,but a set of time se-ries which have the same discrete representation result.The result of this approximate query is a set of time series with the same discrete representation.The advantage of the linear hash index method is that it can perform an approximate query in a lower time complexity.In an approximate query of a time series,the process can be divided into approximate query and result refining.Given a query instance of the time series,return a BSF(Best-so-far)result set through linear hashing,and it can be seen as an approximate query result set of given instance.Finally,get a refine value(RV)set through refining the BSF as a final result.In this process,using a combination method of K nearest neighbors and lower bound distance,which is beneficial to narrow the scope of the query,and realize the refine of the query result.Based on the above theory,the approximate query system of time series is designed and implemented,and the effect of the lower bound is analyzed by using the cardinal,the length of time series.