A foundational framework designed for Intelligent Satellite-based applications incorporating data mining techniques
- Paper ID
20666
- author
- company
Beijing Institute of Satellite Information Engineering, China Academy of Space Technology (CAST);
- country
China
- year
2014
- abstract
During the past decades, numerous advances have accomplished in the field of satellite technologies. Many countries have launched series of satellites with various applied purposes, including U.S., Russia, China, etc., and these satellites have created massive datasets accumulating several terabytes of data each day for scientific applications. As a result, many applications have been raised to process and archive the data sending back from satellites. However, how to efficiently handle the huge amount of data and extract useful information from them remain to be further studied. Since data mining is a fast-growing field and involves many useful techniques, such as statistics, machine learning, pattern recognition, database systems, and information retrieval and so on, it is worth mentioning that many data mining techniques have been adopted to acquire deeper understandings of data within satellite-based applications. Consequently, how to utilize data mining techniques applicable to process data generating by satellites is one of the key problems to design intelligent satellite-based applications. Thus in this paper we firstly proposed detailed studies focusing on the attribute of satellite data and the main data mining techniques respectively. Generally, the most common data sending back from satellites are imaging information, positioning information and variety kinds of monitoring information; and each type of data has obvious distinct property. Considering these properties from the specific data, we analyzed the advantages and disadvantages of various algorithms involved in data mining, such as artificial neural networks, decision trees, support and relevance vector machine, Bayes’ theorem, k-nearest-neighbor classification, clustering and so on. And based on that priori knowledge, a foundational framework for applying data mining techniques into satellite-based applications is brought up in detail, which aims at providing some basic guidance when designing such intelligent applications. The framework is constructed in a bottom-up approach, consisting of data preprocessing, data warehousing, outlier detecting, data clustering, data classifying, frequent patterns mining and association rules generating with regard to specific applications, and etc.. Additionally, some interesting applications based on this foundational framework are outlined with their substantive workflows, such as discovering real-time traffic conditions using massive sequential positioning requests from distinct users. Finally, we summed up the whole paper with a compact description of the foundational framework designed for intelligent satellite-based applications incorporating data mining techniques, and some typical applications based on that framework with their detailed workflows. Key Words: data mining, framework, intelligent satellite-based applications