Conventional, Neural, and Genetic Spatial Interaction Models

Click here to start

Table of Contents

Conventional, Neural, and Genetic Spatial Interaction Models

First Some Introductory Comments

There is an increasingly serious problem caused by IT developments ..

its called ?????????

there are LOTS of it!

Everyday there is More DATA than previously

More Data and More Data and More Data and More Data and More Data and More Data and More Data and More Data and More Data and More Data and More Data and More Data and More Data and More Data and More Data and More Data and More Data and More Data and

No!We now store it in Data Archives also called Data Warehouses and Data marts!

Much of the Information is Spatially Referenced

Much of the DATA are SPATIAL INTERACTION data

Now...you would think that ??

Geographic Information Systems

Where are the tools for Analysis and Modelling that can cope?

A new era that is Spatial Data Rich needs a New Set of Modelling and Analysis Tools to exploit the data

The Re-Birth of Spatial Interaction Modelling reflects the FACT that much of the DATA RICHES are flow data of sort or another

Spatial Interaction Models are:

Creating FLOW data

Flow data aggregation

Geographical AGGREGATION tends to be very useful at drawing out the patterns in databases

Aggregation creates the Flow Data that Spatial Interaction Models can handle

the HARDEST part in Spatial Interaction Modelling is understanding the basic components

An Example of a Flow Matrix

Basics of Flow Tables or Interaction Matrices..

The key inputs to a Spatial Interaction Model are..

The REALLY HARD PART is...

What is a CONSTRAINED model?

Basic Sequence of Steps in Spatial Interaction Modelling

SIMS makes these tasks easy

The traditional route to understanding SI models is via equations

The SIMS allows you to treat the process as a series of Black Boxs

The Modelling Process..

Which model is BEST?

So what can you change to improve a Model ‘s fit to data?

SIMS - Spatial Interaction Modelling System

SIMS .. continued

SIMS .. continued

In the Age of Data Warehouses the new needs are for: Data Driven Modellingvia Inductive Mechanisms

GeoComputation Paradigm

Neurocomputing is very attractive technology because:

Neural Headaches

Black Box Modelling of processes is regarded as unhelpful!

Neurocomputing is now an accepted and widely diffused technology

An illustrative empirical study comparing the performance of: maths models, neural nets, fuzzy logic, and genetically bred models of spatial interaction

Generic Modelling Process

Some Evaluation data

A quick look at FOUR different modelling approaches to Spatial Interaction

Conventional Models

Artificial Neural Networks

Genetic Model Breeder

Model Breeding Results

Fuzzy Logic Spatial Interaction Models

Author: Stan Openshaw

Email: stan@geog.leeds.ac.uk

Home Page: http://www.geog.leeds.ac.uk/staff/s.openshaw/