A hybrid approach based on rough sets theory-decision tree for land use changes monitoring using TM images (Case study: Shushtar City)

With the development of science and technology, the tremendous amount of spatial and non-spatial data have been stored in large data bases. Analyzing these data for decision is seriously in need of spatial data mining and knowledge discovery to provide knowledge. Using satellite images, geo-statistical analysis, and all kind of spatial data are useful and applicable tools in land use changes monitoring; but what’s important among them is how to extract rules throughout big data for knowledge discovery. Rough Set Theory (RST)
is one of data mining techniques which is applied for uncertainty modeling in different ways. Hence, this paper applies an integrated method of RST-DT for satellite images classification and land use changes monitoring. The RST algorithm is used in order to extract reasonable rules.From the results of research, regardless of the changes occurred during three periods (1986-1998, 1998-2014 and 1986-2014) can be found that the land use changes has drastically occurred into residential and water bodies classes by the rate of increasing and decreasing, respectively. Compared to former classes, farm lands have changed a little during the periods. With respect to the base year (1986 or 1365), the area of agricultural lands compared to the base year, coincided with the war period, has shown little changes. This means currently the area of agricultural lands is similar that of during the war. It could be attributed to a tragedy that has been happening in the agricultural sector. Accuracy and Kappa ratio from hybrid model of DT-RST display the RST as a powerful tool in data mining, to reduce superfluous data from the database and extracting rules in order to apply in DT method.