Summary of Abstract Submission

Abstract Submission No. IO50-08-0029PresentationPoster

FILLING OF DATA GAPS IN SEA SURFACE TEMPERATURE TIME SERIES DATA USING HADOOP BASED NEURAL NETWORKS

Jonnakuti Pavan Kumar*1, T V S Uday Bhaskar1, E Pattabi Rama Rao1

1 INCOIS, INDIA

ABSTRACT :

Large scale satellite data are generated continuously by multiple sensors in daily communications. Forecast on such data possesses high significance for analyzing the behaviors of huge amounts of data. However, the natural properties of satellite data present three non-trivial challenges: large data scale leads it difficult to keep both efficiency and accuracy; similar data increases the system load; and noise in the data set is also an important influence factor of the processing result. To resolve the above problems, We can work efficiently with the neural networks on large data sets. Data is divided into separated segments, and learned by a same network structure. Then all weights from the set of networks are integrated and renovate the conventional back propagation neural network to the next layer using big data techniques. A Hadoop based framework called HBNN (i.e. Hadoop-based Back propagation Neural Network) is proposed to process forecast on large-scale SST data. It utilizes a diversity-based algorithm to decrease the computational loads. Extensive experiments on gigabyte of realistic SST data are performed on a Big Data platform and the results show that HBNN is characterized by greater efficiency, good scalability and anti-noise.