Interesting spatiotemporal rules discovery: application to remotely sensed image databases

• Published in: VINE. Very informal newsletter on library automation 2011-01, Vol.41 (2), p.167-191
• Main Authors: Boulila, W, Farah, I.R, Solaiman, B, Ben Ghézala, H
• Format: Article
• Language: English
• Subjects: Analysis; Automation; Computer Science; Content analysis; Criteria; Data collection; Data mining; Decision support systems; Exact sciences and technology; Farms; Grasslands; Image databases; Image Processing; Information and communication sciences; Information and document structure and analysis; Information processing and retrieval; Information science. Documentation; Knowledge discovery; Knowledge management; Land use; Machine Learning; Principal components analysis; Reasoning; Remote sensing; Remote sensing data; Sciences and techniques of general use; Scientific and technological watch. Knowledge management; Studies
• ISSN: 0305-5728; 2059-5891; 1474-1032; 2059-5905
• DOI: 10.1108/03055721111134808

Purpose - Knowledge discovery in databases aims to discover useful and significant information from multiple databases. However, in the remote sensing field, the large size of discovered information makes it hard to manually look for interesting information quickly and easily. The purpose of this paper is to automate the process of identifying interesting spatiotemporal knowledge (expressed as rules).Design methodology approach - The proposed approach is based on case-based reasoning (CBR) process. CBR allows the recognition of useful and interesting rules by simulating a human reasoning process, and combining objective and subjective interestingness measures. It takes advantage of statistics' power from objective criteria and the reliability of subjective criteria. This helps improve the discovery of interesting rules by taking into consideration the different properties of interestingness measures.Findings - The proposed approach combines several interestingness measures with complementary properties to improve the detection of the interesting rules. Based on a CBR process, it, also, offers three main advantages to users in a remote sensing field: automatism, integration of the users' expectations and combination of several interestingness measures while taking into account the reliability of each one. The performance of the proposed approach is evaluated and compared to other approaches using several real-world datasets.Originality value - This study reports a valuable decision support tool for engineers, environmental authority and personnel who want to identify relevant discovered rules. The resulting rules are useful for many fields such as: disaster prevention and monitoring, growth volume and crops on farm or grassland, planting status of agricultural products, and tree distribution of forests.