Search

Typhoon trajectory related data involves many factors such as atmospheric factors, oceanic factors, and physical factors. It has the characteristics of high dimension, strong spatio-temporal correlation, and non-linear correlation, which increases the difficulty of typhoon trajectory prediction. Using feature selection approaches to select appropriate prediction factors becomes an important means to reduce the dimension of typhoon trajectory related data and improve the performance and accuracy of typhoon trajectory prediction methods. However, the existing feature selection methods based on linear correlation analysis can not well depict the nonlinear correlation between data features, which results in low accuracy of feature selection. The feature selection methods based on nonlinear correlation analysis are computationally expensive, which affects the timeliness of feature selection. To solve the problem, we propose a parallel feature selection method NX-Spark-DC based on the Spark platform for typhoon trajectory related data. The method firstly filters out the redundant features of typhoon related data by Normalized Mutual Information (NMI) method, subsequently eliminates the useless features by XGBoost machine learning model, and thus reducing the dimension of typhoon related data. On this basis, an improved Spark-based parallel distance correlation algorithm (Spark-DC) is proposed to select the feature combinations with strong correlation. A series of experimental results show that NX-Spark-DC method has high execution efficiency and accuracy, which is significantly better than the existing methods.

Sea Surface Temperature (SST) is an important physical quantity of the ocean system. The accurate prediction of SST is essential for studying physical ocean phenomena and forecasting the ocean environment information. In this paper, a SST prediction approach based on 3-Diminsional Convolutional Neural Network (3D CNN) and Long Short-Term Memory (LSTM) network with attention mechanism is proposed, which considers the spatial correlation and temporal dependency of SST data. Firstly, the machine learning algorithm XGBoost is used to extract the long period time features of each SST data. Then the 3D CNN is used to capture the spatial correlations among SST field data composed of multiple observation points in a selected sea area, followed by the LSTM model to extract the time dependency features of the SST field time series data, and the attention mechanism is added to weight the output of each step of LSTM model to adjust the prediction results and improve the prediction accuracy of the approach. A series of experimental results show that the proposed approach has lower complexity, higher training efficiency and prediction accuracy, which is significantly better than the existing prediction models.

Data pre-deployment in the HDFS (Hadoop distributed file systems) is more complicated than that in traditional file systems. There are many key issues need to be addressed, such as determining the target location of the data prefetching, the amount of data to be prefetched, the balance between data prefetching services and normal data accesses. Aiming to solve these problems, we employ the characteristics of digital ocean information service flows and propose a deployment scheme which combines input data prefetching with output data oriented storage strategies. The method achieves the parallelism of data preparation and data processing, thereby massively reducing I/O time cost of digital ocean cloud computing platforms when processing multi-source information synergistic tasks. The experimental results show that the scheme has a higher degree of parallelism than traditional Hadoop mechanisms, shortens the waiting time of a running service node, and significantly reduces data access conflicts.

Based on the features of marine environmental data and processing requirements, a cloud computing architecture of marine environment information is proposed, which provides a new cloud technology framework for the integration and sharing of marine environmental information resources. A physical layer, software platform layer and an application layer are illustrated systematically, at the same time, a corresponding solutions for many difficult technical problems such as parallel query processing of multi-dimensional, spatiotemporal information, data slice storage, software service flow customization, analysis, reorganization and so on. A prototype system is developed and many different data-size experiments and a comparative analysis are done based on it. The experiment results show that the cloud platform based on this framework can achieve high performance and scalability when dealing with large-scale marine data.

Aiming at the problems of the spatial join query processing in the cloud environment, an effective spatial join query processing optimization algorithm is proposed in this paper, which is based on the MapReduce framework. The algorithm applies a grid partitioning method to distribute spatial data objects and uses a boundary filtering strategy to reduce the computation and communication cost of spatial join query processing. Firstly, data space is divided into partition units with the same size, and spatial data objects are distributed into the corresponding partition units according to the inclusion relation between them. Then calculates the MBR of the spatial data objects within a partition unit and uses the MBR to filter the useless data objects across the partition unit, thereby reducing the corresponding computation cost. At the same time, a simple and effective data results duplication avoiding mechanism is applied to avoid repeating output of spatial join processing results, further reducing the computation cost at Reduce stage. Experiment results on synthetic and real datasets show that the proposed algorithm has obvious advantages and good performance than the original SJMR query processing algorithm.

Due to having many advantages, virtualization technology has been widely used and become a key technique of cloud computing. Live migration of virtual machines is the core and key technique of virtualization fields, but the existing pre-copy live migration approach has the problems of low copy efficiency and long total migration time, so we propose an extreme learning machine (ELM) based adaptive live migration approach of virtual machines (ELMBALMA) in this chapter. Firstly, the approach uses the ELM algorithm to classify the virtual machines according to the type of the running applications, and then choose the best suitable migration algorithms for each type of virtual machines, thereby reduce the time of live migrating of virtual machines. In addition, we proposed a memory compression based live migration algorithm (MCBLMA) for the memory-intensive application scene. The algorithm uses a weight-based measurement method of writable working set, which can accurately measure the writable working set, so that it can reduce the amount of dirty memory page transmission, meanwhile it uses a memory compression algorithm to compress memory pages to be transmitted, and thus reduces the data transmission time. Preliminary experiments show that the proposed approach can significantly reduce the memory pages transmitted, the total migration time and the downtime of virtual machines.

Although P2P systems are widely used for sharing resources and retrieving information among numerous users, one of the fundamental challenges faced by P2P systems is to construct an overlay network, which can efficiently process complex queries on high-dimensional data. In this paper, we propose a Semantic Cluster Based Overlay structure (SCBO), in which peers are clustered into semantic clusters according to their data semantic information, and then these clusters are formed into a semantic overlay network. And an efficient KNN query processing algorithm is proposed for high-dimensional data objects based on SCBO structure. Experiments on real and synthetic data set demonstrate that SCBO and the proposed KNN query algorithms have well performance and scalability.