Design of Efficient Mining Methods for Spatial Association Rules

• Main Authors: Mei-Hsiu Chen, 陳美秀
• Other Authors: Chin-Feng Lee
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/17220816157639462918

碩士 === 朝陽科技大學 === 資訊管理系碩士班 === 91 === Since huge amounts of spatial data can be easily collected from various applications, ranging from remote sensing technology to geographical information system, the extraction and comprehension of spatial knowledge is a more and more important task. Spatial association rule mining is a kind of spatial data mining that is to discover interesting and implicit association knowledge from spatial databases. There have been some interesting studies related to the mining of spatial association rules. However, lack of studies on semantic spatial association rules which can reflect the way human think. Besides, designing efficient methods for mining spatial association rules also demand immediate our attentions from better performance. Therefore, in this thesis, an efficient method based on Class Inheritance Tree (CIT) is proposed for capturing intrinsic relationships between spatial and non-spatial data. This rules help to accommodate data semantics as well as to achieve better performance. Moreover, many excellent studies on Remote Sensed Image (RSI) have been conducted for potential relationships of crop yield. However, most of them suffer from the performance problem because their techniques for mining association rules are based on Apriori algorithm. In this thesis, two efficient algorithms, two-phase spatial association rules mining and adaptive two-phase spatial association rules mining, are proposed for addressing the above problem. Both methods primarily conduct two phase algorithms by creating Histogram Generators for fast generating coarse-grained spatial association rules, and further mining the fine-grained spatial association rules w.r.t the coarse-grained frequently patterns obtained in the first phase. Adaptive two-phase spatial association rules mining method conducts the idea of partition on an image for efficiently filtering out non-frequent patterns and thus facilitate the following two phase process. Such two-phase approaches save much computations and will be shown by lots of experimental results in the thesis.