عنوان مقاله [English]
The most important challenge in urban land use planning is the spatial organization of urban activities and functions based on the needs of urban society. Extracting the current rules that exist in the city not only adds local conditions to the standard values mentioned in various instructions (Habib et al, 1999; Shiea 2018; Saeedinia 2004) but also makes it possible to analyze with comparing the existing conditions of the city with the standards. There is some research to examine the current situation of the city. Most of these studies have used statistical methods (Hosseinzadeh et al. 1399; Omidipour et al, 2017; Mohammadnejad et al. 2012).
A few of them have utilized data mining methods, but none of these studies examine existing patterns between one type of land use with other land uses. In addition, the method used in this research is a new method that tries to use the capabilities of association rules and decision trees in exploring co-located patterns by combining these methods.Therefore, considering the importance and necessity of addressing this issue, the purpose of this research is to explore the current situation of urban land use by using data mining methods to discover the current patterns in the location of land uses in the vicinity and at different distances.
Finally, providing rules derived from these models may help planners and managers to understand the current status of land use appropriately and improve urban land-use plans by utilizing them in combination with standards and rules based on expert knowledge.
Materials & Methods
Spatial association rules
Association rules discover the laws of interdependence between the data of a large database. In other words, patterns that are frequently repeated in the data set are identified and used to explain the rules of dependence (Han & et al, 2011: 54; Li 2015). The rules of the association in which one of the propositions in the premise or sequence contains a spatial relation are called spatial association rules (Geissen & et al, 2007: 277-287, Mennis & et al, 2005: 5-17).
The decision tree is one of the most powerful and common techniques for classification and prediction. Among the algorithms used to construct the decision tree, the most important is the C5 algorithm which is the developed ID3 algorithm.
A n*l transaction matrix is generated. Where n is the number of available features and l represents the number of types of land use studied, which is 19 in this article. The elements of this matrix can be zero or one.
To fill the transaction matrix, we first consider the distance and apply buffer analysis for all the features in the land use layer. Then, for each feature, we intersect the buffer layer of that feature with the land-use layer and extract all the features that appeared at the intersection. Arc GIS software was used to perform spatial analysis.
Then, to extract the current rules of land use in the urban environment, the a priori algorithm is selected as one of the association rules algorithms, and the C5 algorithm is selected as one of the decision tree algorithms.
In this research, the user data of neighborhood 4, district 5 of Tehran Municipality, including 1065 property plots, were used.
Results & Discussion
In this step, the proposed model for deriving the rules of land use dependence based on the current situation of land use in the study area is implemented step by step and the results are presented.
According to existing standards, three distances are considered to extract spatial rules with an apriori algorithm. After extracting the rules, they are compared with the values of approved standards in urban land use planning. Vicinity and compatibility are examples of indicators in common standards for locating and determining land use for the land. Using the extracted rules, the indicators are examined.
Due to the lack of extraction of some rules by association rules, for example, not extracted rules related to therapeutic land uses within 300 meters from residential land uses, we use the decision tree algorithm to extract related rules in more detail. The graphs obtain from the decision tree shows which land uses are effective for predicting and categorizing specific land uses, based on the current status of the land uses located in the case study area.
The purpose of this paper is to data mining the current status of urban land uses to extract the rules of neighborhood and proximity of different land uses. Using the proposed model in this article, it is possible to extract the existing rules of land uses in detail and as well as to evaluate its compliance with conventional standards and criteria in urban land use planning.