Principles of Neural Spatial Interaction Modelling
نویسنده
چکیده
The focus of this paper is on the neural network approach to modelling origin-destination flows across geographic space. The novelty about neural spatial interaction models lies in their ability to model non-linear processes between spatial flows and their determinants, with few – if any – a priori assumptions of the data generating process. The paper draws attention to models based on the theory of feedforward networks with a single hidden layer, and discusses some important issues that are central for successful application development. The scope is limited to feedforward neural spatial interaction models that have gained increasing attention in recent years. It is argued that failures in applications can usually be attributed to inadequate learning and/or inadequate complexity of the network model. Parameter estimation and a suitably chosen number of hidden units are, thus, of crucial importance for the success of real world applications. The paper views network learning as an optimization problem, describes various learning procedures, provides insights into current best practice to optimize complexity and suggests the use of the bootstrap pairs approach to evaluate the model’s generalization performance.
منابع مشابه
Computational neural networks: a new paradigm for spatial analysis
In this paper a systematic introduction to computational neural network models is given in order to help spatial analysts learn about this exciting new field. The power of computational neural networks viz-a-viz conventional modelling is illustrated for an application field with noisy data of limited record length: spatial interaction modelling of telecommunication data in Austria. The computat...
متن کاملRainfall-runoff modelling using artificial neural networks (ANNs): modelling and understanding
In recent years, artificial neural networks (ANNs) have become one of the most promising tools in order to model complex hydrological processes such as the rainfall-runoff process. In many studies, ANNs have demonstrated superior results compared to alternative methods. ANNs are able to map underlying relationship between input and output data without prior understanding of the process under in...
متن کاملشریعت - شهر اسلامی
Although in the structural approaches to the concept of place identity, various aspects of the built environment, natural environment, social environment and human activities are considered as the factors of the city identifying. In the normative approaches, adaptation of the socio-spatial structure of the city with the values, culture and human needs have major contribution in its identity. Wi...
متن کاملSPATIOTEMPORAL DYNAMIC OF TOXIN PRODUCING PHYTOPLANKTON (TPP)-ZOOPLANKTON INTERACTION
The present paper deals with a toxin producing phytoplankton (TPP)-zooplankton interaction in spatial environment in thecontext of phytoplankton bloom. In the absence of diffusion the stability of the given system in terms of co-existence and hopf bifurcation has been discussed. After that TPP-zooplankton interaction is considered in spatiotemporal domain by assuming self diffusion in both popu...
متن کامل