Development of spatial decision support systems with machine learning techniques: Case of Aksaray province

All every spheres of life, numerous decisions are taken on a macro and micro scale. As the number of criteria that have an influence on the decisions increases, ranking of criteria and calculation of weight values are becoming more complex. Particularly, the decision analysis associated with spatial will be more complex due to data density as well as excess of criteria. For the solution of this issue, Spatial Decision Support Systems (SDSS) methods have been developed based on Geographic Information Systems (GIS) and Decision Support Systems (DSS). Recently, GIS-based Multi-criteria Decision Making Methods have been used in many applications such as selection of appropriate settlement area, property evaluation and landslide susceptibility by the public, private and academic circles. Generally in these practices, methods requiring expert opinion are preferred and accepted. In these expert-based methods, factors such as differences in views of experts and subjective evaluation, caused by interpretation make it difficult to popularize and automate applications. Furthermore, dynamic models cannot be created as the spatial analysis needs to be revised when the changes (criteria adding / decrease) that will occur in the data structure. Dynamic methods need to be developed in order to minimize the human impact in decision analysis so that decision making can be fast, effective and objective. In this context, artificial intelligence techniques, which are increasingly used in many different applications, need to be integrate SDSS. Over recent years self-learning systems have been developed with the advances in Artificial intelligence techniques. In this study, it is aimed to integrate machine learning techniques of artificial intelligence into the data driven SDSS and automate these systems. For this purpose, supervised learning algorithms such as Artificial Neural Networks, Support Vector Machines, CHAID, ID3, C4.5, CART and Random Forest were used for the generation of the Aksaray Province real estate valuation map and their performances were evaluated. In addition, the results obtained were compared by using the statistical method, Multiple Regression Analysis, which is frequently used in the literature. A software covering all the methods mentioned above was developed to enhance.