Spatial Data Mining and Data Analysis

With the advance growing of technology, expansion of research areas, deployment of different commercial and open sources GIS systems has lead out a massive collection of data stored in different debases. Nowadays, we generate about several trillion bytes of data every day, characterized by high dimensionality and large sample size and called Big Data or massive volumes of data. However, in today’s situation the data is mysterious, we have data rich but information poor.DM is the non-trivial process of identifying valid, novel, potentially useful and ultimately understandable patterns in data. Fayyad et al. (1996)

Spatial data mining in other way is a distinctive kind of data mining. The main distinction between data mining and spatial data mining is that, in spatial data mining tasks we use not only non-spatial attributes but also spatial attributes.It has been said that, the spatial data are special, and therefore treating or analyzing them needed special methods and techniques. This concept has appeared in different review papers and articles although few of them are against on this notion.Most of this review papers suggests that it is tremendous difficult task to mines interesting patterns in a geographical datasets compared to extracting them in traditional data, this is because geographical or spatial data is associated with complex spatial data types, spatial relationship, spatial heterogeneity, spatial autocorrelation, ecological Fallacy and Modifiable Areal Unit Problem (MAUP) In a such kind of situation, the adoption and effectiveness of traditional data mining techniques become thankless. In order to decide whether the spatial data are special or not, I suggest to spend our little time to have short tour to describe the term spatial analysis and then describe only two characteristics of spatial data.

Spatial analysis is special kind of methods with the aim of identifying or describing the pattern to identifying and understanding the process associated with that particular pattern. The results of Spatial analysis change when the locations of the objects being analyzed change. This is well explained by Tobler’s (1979) in his First law of Geography "everything is related to everything else, but near things are more related than distant things." The first law of geography is more emphasizing on spatial dependency or spatial autocorrelation which implies that the phenomenon on one location are more likely to repeat in a location near to it than its far away. Dealing with this kind of situation a very special techniques are needed First to compare the observed pattern in the data (e.g., locations in point pattern analysis, values at locations in spatial autocorrelation) to the one in which space is irrelevant (Anselin, 1989)

These datasets are scale dependent, query associated to extract information for this data set are more advanced and much complex as explained in . This oppose with the traditional statistics techniques which assume that observations are independent, and therefore in this sense these techniques cannot be implemented critically for the data showing spatial dependency behavior. Spatial data has another unique features called spatial heterogeneity which means that the behavior on relationship over space are not stable, they vary in different areas of the map. A realistic perspective on most spatial data has to assume that in general most spatial processes are nonstationary and anisotropic. Heterogeneity and nonstationary create additional problems in analysis, emphasizing the local nature of space/process interaction”.

In my point of view based on a number of readings, if we take an aerial sight on this reviews, there is no doubt that both of their arguments has one theme in common,” the tradition statistical technique cannot be applied on analyzing spatial data, the reasons behind this, is that; spatial data has a specific characteristic which make them to be more complex in developing a coherent and sturdy approach to spatial analysis based on traditional statistical techniques  .”Although in their review paper Marco Painho argue that “special nature of spatial data has lost and their fore spatial data has the same characteristics as a secondary data ”. But they were clear that, advance methods and technologies are needed to analyze these kind of datasets due to their unique features. Unique feature of spatial data is still emerged, the demands of GIS Science to perform advance analysis for accurate or error free become larger and complexity and need to understand spatial become a major reason for the need of GEOSPATIAL DATA MINING, as a central discipline to deal with modeling and analyzing spatial data.

List of References

1. Anselin, L. (1989, 4 1). What is Special About Spatial Data? Alternative Perspectives on Spatial Data Analysis (89-4). Retrieved from https://escholarship.org/uc/item/3ph5k0d4: https://escholarship.org/uc/item/3ph5k0d4
2. Mark Gahegan, H. M. (n.d.). Geospatial Data Mining and Knowledge Discovery. Retrieved from https://pdfs.semanticscholar.org/7e09/92a048952b4ee61b2993c4412b9db117ac97.pdf
3. Painho M, B. F. (2005). ON THE PARTICULAR CHARACTERISTICS OF SPATIAL DATA AND ITS SIMILARITIES TO SECONDARY DATA USED IN DATA MINING. Retrieved from www.novaims.unl.pt: http://www.novaims.unl.pt/docentes/vlobo/Publicacoes/2_13_lobo05_secondary_data.pdf
4. Ramón Giraldo, S. D.-N. (2018). Statistical modeling of spatial big data: An approach from a functional data analysis perspective. SCIENCE DIRECT, 1.
5. Williamson, I. (2009, June). Is Spatial Special? Retrieved from [email protected]: http://www.csdila.unimelb.edu.au/publication/misc/is-spatial-Special-Ian-june09.pdf