In this chapter, the authors propose an approach to predict uncertain spatiotemporal data. This approach is unique in the predicting element nodes which are integrated into the position element node in uncertain spatiotemporal XML data tree. At the same time, the other element nodes do not need to make any changes. In addition, the authors apply this method to meteorological applications and established a series of experimental models for testing. PGX (predictive model with grey model based on XML), which is applied to uncertain spatiotemporal objects, is able to achieve the minimum mean accuracy of 0.5% in a short time. The experimental results show that PGX can effectively improve the efficiency of information storage and retrieval. The experimental prediction accuracy is guaranteed (the relative error is between 0.5% and 5%) and the query time based on XML is 89.2% shorter than that of SQL Server.
Top1 Introduction
Spatiotemporal data are featured by spatial and temporal phenomena (Santos et al., 2015). Previous efforts started with separate research in temporal and spatial database areas. Along with the development of Mobile Devices and the advance of geographic information systems, a great deal of geoscientific data has been generated. These data lie in continuous space and change with time and space, so they have a high degree of particularity and complexity (Jeon et al., 2010). Hence, spatiotemporal data management is becoming more and more important, especially spatiotemporal data prediction.
The spatial and temporal uncertainty exists widely in the complex and variational real world. For instance, the incompleteness of knowledge, the fuzziness of concepts and the derivation of data make the entity possess the feature of uncertainty. Therefore, uncertainty is absolute while certainty is relative.
Recently, studies on predicting uncertain spatiotemporal objects have made great progress (Tao et al., 2004). Jeung et al. (Jeung et al., 2008) presented an approach which forecasted an object’s future locations in a hybrid manner by using motion function and movement patterns. Cheng et al. (Cheng & Wang, 2008) used a dynamic recurrent neural network for spatial prediction, which is applied to forest fire. The trajectory of moving target, which is limited by traffic network, can be predicted accurately. (Jeung et al., 2010). The work of Boulila et al. (Boulila et al., 2011) takes imperfection into account, relating to the spatiotemporal mining process in satellite imaging. The particularity and complexity of the spatiotemporal data are enhanced. Because spatiotemporal data have been applied increasingly and widely in the field of transportation, meteorology, earthquake rescue, criminal analysis, web applications, public health and medical services (Xu et al., 2016). Therefore, it is urgent to improve the predicted technology of uncertain spatiotemporal data.
Coz et al. (Le Coz et al., 2014) developed a method with a statistical representation of uncertainties. In (McMillan et al., 2010), McMillan et al. provided a comprehensive review of the uncertain values for gauging and rating curves. However, in the field of XML, there are few researches on the prediction of uncertain spatiotemporal data. Additionally, compared with the certain spatiotemporal data, uncertainty enhances spatiotemporal data particularity and complexity, making it more challenging to predict uncertain spatiotemporal data accurately. Therefore, a new modeling technology of predicting uncertain spatiotemporal data is required. Researches on grey dynamic model have been investigated widely in recent years. Original data in grey dynamic model are composed of both certain and uncertain information. By obtaining part of the data, the correlation of all factors is measured, the data are preprocessed, and finally the differences and similarities of the development trend of all factors are obtained. The predicted sequence can be obtained by processing the current information. Meanwhile, the predicted results are accepted when the accuracy of the predicted meets the reliability requirement. In this case, the grey dynamic model can accurately predict the changing trend of the uncertain state of the spatiotemporal data. Therefore, the grey dynamic model is integrated into the prediction of uncertain spatiotemporal data. There is a series of efforts to improve the perception of the grey theory (Liu et al., 2012; Yang et al., 2014). In (Hamzacebi & Es, 2014), Hamzacebi and Es show the superiority of Optimized Grey Modeling (1, 1) when compared with the results in literature. The grey relational analysis of the fuzzy sets also takes an important occupation in the fuzzy measure field (Sun et al., 2018). Furthermore, Zhang et al. (Zhang et al., 2013) propose a grey relational projection method for the MADM problems with intuitionistic trapezoidal fuzzy number attribute. However, the grey theory is studied on combination with other techniques rather than itself (Kuang et al., 2015; Yang et al., 2014). In addition, it has been applied to other applications such as the electricity consumption. Actually, the theoretical basis of grey dynamic model is more beneficial to predict uncertain spatiotemporal data (Bai et al., 2017).