22
Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. Louis Gary Raines Department of Computer Science US Geological Survey [email protected] [email protected] v

Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

Embed Size (px)

DESCRIPTION

What is the problem?  Project mineral-related activity on public land to 2010 Predicting permit activity in an area Spatially explicit USGS and others have data on permit activity from 1989 – 1998 as well as data on natural resources Use cellular automata to model (predict) mineral activity over next ten years Problem: Takes weeks to tune CA rules to match available data

Citation preview

Page 1: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

Predicting permit activity with cellular automata calibrated with

genetic algorithms

Sushil J. Louis Gary Raines

Department of Computer Science

US Geological Survey

[email protected] [email protected]

Page 2: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Outline

What is the problem? Calibrating a CA

What is the technique? Genetic Algorithm

What are the issues? Discretization Encoding Evaluation

What are our results ?

Page 3: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

What is the problem?

Project mineral-related activity on public land to 2010 Predicting permit activity in an area

Spatially explicit USGS and others have data on permit activity from 1989 – 1998

as well as data on natural resources Use cellular automata to model (predict) mineral activity over

next ten years Problem: Takes weeks to tune CA rules to match

available data

Page 4: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

What is the problem?

Can we automate calibrating a cellular automaton As good as CA calibrated by human In the same or less time

Page 5: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

What is the problem?

Page 6: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Model calibration as search

Search through the space of possible model parameters to find a parameter set that fits observed data

Many search methods We use genetic algorithms

Page 7: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Genetic Algorithms

Poorly understood problems (Holland, ‘75, Goldberg, ‘89) Empirical evidence to support their use in this kind of

problem Physics models

Physical Review Letters, Volume 88, Issue 4 Journal of Quantitative Spectroscopy and Radiative Transfer. Volume

75, 2002, Pgs. 625 - 636 Seismic models

Congress on Evolutionary Computing 1999, pages 855 - 861 Hydrology models

In progress Proceedings of GECCO, CEC, …

Page 8: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Genetic algorithm calibration

Page 9: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

What is a GA?

Randomized, parallel search Models natural selection Population based Uses fitness to guide search

Page 10: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Genetic algorithm search

Page 11: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Genetic Algorithm

Randomly initialize P(0) with candidate parameter sets

Loop Select P(t+1) from P(t) Crossover and Mutate P(t+1) Evaluate P(t+1) run CA model t = t+1

Page 12: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Modified Annealed Voting Rule Probability of Life in Next Generation

Number of Live NeighborsStatus of Center Cell

Alive Dead

> Annealing Window Very Likely LikelyAnnealing Window Likely Somewhat

Likely< Annealing Window Very

Somewhat Likely

Unlikely

Page 13: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Definitions of Parameters

Parameters DefinitionVery Likely Square root of Likely (Larger)Likely A high probability of life.Somewhat Likely An intermediate probability of lifeVery Somewhat Likely Square root of Somewhat Likely (Larger)

Unlikely A low probability of lifeResource Threshold Minimum fuzzy membership defining where

a reasonable explorationist would exploreAnneal Window Position and width control response of CA

Page 14: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

GA Encoding

GA usually works with string structures representing a candidate solution

2^36 = 64Gig possibilities Fitness = scaled match to observed data

top bottom likely slikely unlikely rt4 4 7 7 7 7

Page 15: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

GA Parameters

Population sizes – 50 Elitist selection – next generation is best of

parents and offspring Probability of crossover – 1.00 Probability of mutation - 0.05 Fitness scaling – 1.05

Page 16: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Model parameters

496 X 503 = 249,488 cell CA 4 or 5 years (iterations) Average over 3 runs Cell data imported from GIS

Page 17: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Results

Page 18: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Results

Page 19: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Results

Page 20: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Results

Page 21: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

GA produces good parameter values (20% better than human)

GA is a viable tool for model exploration

Many different parameter sets give about the same fit ?

Modeling rare events ?

Page 22: Predicting permit activity with cellular automata calibrated with genetic algorithms Sushil J. LouisGary Raines Department of Computer Science US Geological

http://gaslab.cs.unr.edu

Cross-Tabulation 1989-1998Number of

CellsCA Trace

0 1 2 3 4 5 6 7 Sum

ActualTrace

0 66364 1671 267 176 50 7 11 0 68446

1 354 136 76 70 24 4 2 0 666

2 154 42 57 49 18 4 1 0 325

3 129 69 102 133 47 29 20 3 532

4 33 32 52 78 34 20 16 1 266

5 8 11 15 25 42 31 11 5 148

6 8 4 22 34 24 22 14 3 131

7 17 4 17 34 81 125 70 25 373

Sum 66967 1969 608 599 320 242 145 37 70887