ESSE 3600 Lecture Notes - Lecture 11: Stochastic Process, Random Field, Inductive Reasoning
DepartmentEarth, Space Science and Engineering
Course CodeESSE 3600
This preview shows half of the first page. to view the full 2 pages of the document.
GIS Lecture 11
Equal interval classification-note that there are three categories to define classical attributes
Random field means each location is XYZ
-Field data is treated as a random variable
Clustering means to group a dataset into clusters based on their principle, maximizing interclusteral
similarity and minimizing interclusteral similarity
-ABC is defined according to our parameters
-Spatial entities have similar properties and other spatial entitites have nonsimilar properties
-Random errors help us analyze similarity or dissimilarity
-We can use different classifictation algorithms
-The proximity measure measures similairy in Gearys index
-There are certain relationships depending on proxminty
-Nearest neighbor method involves finding what is the closest
-Spatial patterns are related to spatial location
-IDW-the weight is the inverse of distance bw red & green
-Structural vs geographical matrix
-GIS is dependent on a process
-The object has to be described.
-Attributes help us evaluate the particular object
-binary model, regression model
-A descriptive model describes existing conditions
-prescriptive model is once there is a system, several parameters can be introduced. It is a mathematical
system that allows us to predict for example, vegetation maps. The system is built and is good for spatial
-Resource allocation is a model in which people adopt a stochastic model
-Stochastic model helps us to allocate resources
-Deductive models are top down models. ie you already have the theory and suggest a hypothesis from
this. Inductive is the opposite. You want to go from the specific to the general(think deductive and
-ex: i predict it takes 40 minutes to get to YorkU, based on what information i have.
-Inductive model is bottom up, deductive is top down model.
-GIS research is done from an inductive model, trying to generalize.
-Binary model vs index model. Index model is based on raster data.
-Calibration is when arbitrary posiions are selected(refer to hospital example)
-Validation(remember hydrology) is to find out whether the model makes sense
-Cross validation, some portion of th sample is reserved for testing
-We design the sample. Model validation is to judge whether the model is acceptable or not.
-Sensitivity analysis is to make sure the attributes #1-3, and all attributes are summed. SA is a technique
that can quantify the model uncertainties by measuring the effects of input changes on the output. Ex:
vary one variable and see its effect on the markets[stock market example]
-Presenting decision to customers, weight values are changed. Within certain ranges, decision doesnt
You're Reading a Preview
Unlock to view full version