Output details
13 - Electrical and Electronic Engineering, Metallurgy and Materials
University of Surrey
Information Abstraction for Heterogeneous Real World Internet Data
This paper describes using time-series analysis, probabilistic machine learning and logical reasoning techniques for processing and interpreting large-scale sensory data. The work addresses one of the key challenges of dealing with world big data sources that provide real world observation and measurement data. A pattern construction method with sliding windows for real world time-series data is developed that creates string patterns from raw sensory data. Results show significant decrease in size of data that is required to be communicated and great potential for large scale and real time event processing and data analysis in dealing with real world big data.