Research On Time Series Data Modeling And Storage

Time Series which has been observed in accordance with the chronological order is an important unstructured data type Relational database is the most important structured data storage and management system. Time Series must be described by a structured data model so that it can be stored in the relational database.Based on extensive research and analysis on time series'characteristics, it is proposed that a structured time series modeling method. This method introduces feature property to describe some facet of the characteristics'and feature type to describe the every facet of the time series'characteristics. Furthermore, combination and inheritance is defined so that old feature type can be reused more easily. A feature description language base on XML is used to represent the feature description. Unstructured time series is represented to a form that contains several sub time series which comprises the feature information by proposed method. Consequently unstructured time series is modeling into structured data. A time series modeling framework is designed and implemented according to the method so that it can modeling various time series came from various field.A rational database storage solution is proposed to store the typed time series and the feature information produced during the modeling procedure. In addition, a prototype is designed and implemented that provides visual query based on the feature of the time series.