پديد آورندگان :
باقري، ميلاد دانشگاه تهران - دانشكده جغرافيا , جلوخاني نياركي، محمدرضا دانشگاه تهران - دانشكده جغرافيا , چاركانه، عبدالخالق دانشگاه تهران - دانشكده جغرافيا , باقري، كيوان دانشگاه تهران - دانشكده جغرافيا
كليدواژه :
MLP , طبقه بندي , شبكة عصبي , توسعة شهري , پتانسيل يابي
چكيده فارسي :
رشد سريع شهرنشيني و توسعة شهري به‌ويژه در كشورهاي درحال‌توسعه، به درك الگو و فرايندهاي پيچيدة رشد شهري با روش علمي و كارآمد نياز دارد. لازمة ايجاد رشد شهري پايدار و برنامه‌ريزي توسعة شهري، درك الگوهاي صحيح رشد شهري است. كرمانشاه نهمين شهر پرجمعيت كشور و يكي از چهار شهر نخست ايران از نظر حادبودن معضل حاشيه‌نشيني است. هدف اين پژوهش، بررسي پتانسيل توسعة شهري در اين شهر است. بدينمنظور، شبكة عصبي پرسپترون چندلايه (MLP) با الگوريتم آموزش لونبرگ- ماركوات به‌كار رفت و داده‌هاي مؤثر در توسعة شهري به‌عنوان لايه‌هاي ورودي به شبكه تعيين شد. اين لايهها كه در سه گروه اجتماعي-اقتصادي، كاربري زمين و بيوفيزيك قرار مي‌گيرند، شامل 16 لايه هستند. در ادامه، پانصد نقطه به‌عنوان نقاط آموزشي شبكه تهيه و 12 لاية مياني نيز تعيين شدند. مطابق نتايج، با دورشدن از امكانات و مناطق شهري، پتانسيل‌ها به‌شدت كاهش پيدا مي‌كنند و بيشتر مناطق داراي پتانسيل توسعة شهري در نزديكترين فاصلة اين امكانات و مناطق شهري قرار دارند. بيشترين مناطق پتانسيل‌دار توسعة شهري، در جنوب‌غرب شهر كرمانشاه و در اطراف جاده‌هاي اصلي كرمانشاه-اسلام‌آباد و كرمانشاه-كنگاور واقع است. مناطق شمالي شهر به‌دليل ارتفاع و شيب زياد، پتانسيل اندكي براي توسعه دارند. همچنين ضريب رگرسيون كلي 95 درصدي شبكه كه حاصل شركت تمامي داده‌ها در شبكه است، كارايي زياد شبكة عصبي پرسپترون چندلايه را در اين مطالعه نشان مي‌دهد.
چكيده لاتين :
Introduction
Some of the important phenomena occurred in recent centuries in the social and economic life of different countries of the world are the emergence of numerous new cities, development of ancient cities, advances in urbanization and urban development. Urban development and changes in land use patterns lead to widespread social and environmental impacts including reduction in natural spaces, increases in vehicle accumulation, reduction in agricultural fertile lands with high production potential and degradation of water quality. Urban development in any country is not coincident with other aspects and, on the other hand, controlling future development requires careful planning. Understanding the right patterns of urban growth is needed to manage sustainable urban growth and planning for urban development. The high rates of urban population growth in Iran and the lack of urban infrastructure in one hand and the increasing trend of land use change, followed by the loss of valuable ecological landuses in urban and peri-urban areas andindustrial pollution provide the necessity of modeling for urban development.
Methodology
The data used in this research can be generally divided into two main categories: the data used to extract land uses in the study area, e.g., satellite imagery, and the data considered as effective factors on urban expansion and land use change. Identifying the variables affecting the creation of the main prerequisites for the development of land use models, we try to use independent groups of variables including socioeconomic, biophysical and land use in this study. Since there are several decision-making rules for exploiting these variables, in this study, the distance between these variables was considered as an indicator. To work with the artificial neural network, the effective parameters in urban development should be given as input to the network (INPUT), and then a number of educational points are provided to the network for using these points (TARGET) to measure the impact of each. It determines the input layers, in fact, the network has learned the necessary training to deal with new areas. After determining the number of hidden layers in the network structure, the entire study area is provided to train the network for zonnation of the potential areas of urban development. MLP network with 16 input layers (effective factors in urban development), 12 intermediate layers (test and error method), a neuron in the output layer as an outline map (final map of urban development potential), and the Leuvenberg Marquard training algorithm was executed and, thus, the network was trained with new samples.
Results and discussion
The network stopped after 44 repetitions and got the necessary training. In repetition 38, the optimal possible condition has the highest correlation and the least error. Given the 0.93 second factor, it can be ensured that network over-network learning is well prevented. Finally, the total regression coefficient of the network is resulted from the company total data network (95%). Then, the entire study area was evaluated by the network and the network based on the weight of the criteria received the training. The output from this stage was a valuable layer between zero and one.
Conclusion
It can be clearly seen that with the evolution of urban facilities and potentials, the vast majority of regions with urban development potential are located at the closest distance between these facilities and urban areas, especially the main roads of the city. The areas displayed in blue have the greatest potential for urban development. The most potential areas of urban development are located in the southwest of Kermanshah and around the main roads of Kermanshah-Islamabad and Kermanshah-Kangavar. The northern areas of the city have a low development potential due to their height and slope. The results of the study could be used to identify the areas for urban development to prevent development of irregular towns that have severe impacts on urban ecosystems and the lives of urban residents, as well as the loss of city capital. The future planning of the city of Kermanshah is relying on sustainable development with the least damage and inconsistencies in the agenda of urban managers and planners.
