Using Neural Networks in Decision Support Systems

clangedbivalveΤεχνίτη Νοημοσύνη και Ρομποτική

19 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

80 εμφανίσεις

P
-
1

© 2005 NeuralWare. All rights reserved.

Using Neural Networks in Decision Support Systems

Introduction


Core
Technology


Building and
Deploying
Neural
Networks






Medical
Procedure
Certification


Crime
Forecasting


Grain Quality
Assessment

Jack Copper

NeuralWare

jack.copper@neuralware.com


Hiroshi Maruyama

SET Software Co. Ltd.


mal@setsw.co.jp

April 2005

P
-
2

© 2005 NeuralWare. All rights reserved.

NeuralWare


Since 1987, NeuralWare has created and marketed neural
network based Artificial Intelligence (AI) software for





Data Mining (clustering)



Classification



Forecasting


NeuralWare collaborates with Customers and Partners to
Embed Intelligent Neural Network Engines into Next
-
Generation Products and Systems

P
-
3

© 2005 NeuralWare. All rights reserved.

Introduction




Characteristics of Neural Network Decision Support Systems


Integrate Data and Analytics


Adapt to Changing Conditions

P
-
4

© 2005 NeuralWare. All rights reserved.

Introduction





Benefits of Neural Network Decision Support Systems


Consistent Decisions


Rapid Decisions


Reproducible Decisions

P
-
5

© 2005 NeuralWare. All rights reserved.

Introduction




Examples of Neural Network Decision Support Systems


Medical Procedure Certification


Crime Forecasting


Grain Quality Assessments


P
-
6

© 2005 NeuralWare. All rights reserved.

Core Technology
-

Neural Networks

Historic Data

Target

Model

Target

Model

Input Layer

Hidden Layer

Output Layer

Decisions Based on Model Output

New Data

Artificial Neural Networks are connected hierarchies
of Artificial Neurons (also called Processing Elements)

P
-
7

© 2005 NeuralWare. All rights reserved.

Building Neural Networks

P
-
8

© 2005 NeuralWare. All rights reserved.

Evaluating Neural Network Performance

P
-
9

© 2005 NeuralWare. All rights reserved.

Evaluating Neural Network Performance

P
-
10

© 2005 NeuralWare. All rights reserved.

Deploying Neural Networks

NeuralWare Technology (Run
-
Time
Engine/Models/FlashCode) embedded
in Server

Browser
-
based wired or
wireless remote PC clients do
not employ

NeuralWare
technology

Server Contains Development and Run
-
Time Engine

Application Server Architecture

P
-
11

© 2005 NeuralWare. All rights reserved.

Deploying Neural Networks

Wired or wireless remote PC
clients employ

embedded
NeuralWare technology (Run
-
Time Engine/Models/FlashCode)

Server Contains Development Engine

Distributed Intelligence Architecture

P
-
12

© 2005 NeuralWare. All rights reserved.

Case Study


Medical Procedure Certification

Objectives


Reduce Workload on Doctors and Registered Nurses


Improve Responsiveness to Customers (faster decisions)

Challenges


No “Gold Standard” for decisions


even Doctors sometimes disagree


Inconsistent data formats and labeling

Process


Used NeuralSight to build and evaluate ~ 30,000 Models in 3 weeks


Developed prototype software to permit altering Model decision threshold


P
-
13

© 2005 NeuralWare. All rights reserved.

Case Study


Medical Procedure Certification

Performance of best models
(ranked by Average
Classification Rate) for the
Global model and CT and MRI
Modality models

P
-
14

© 2005 NeuralWare. All rights reserved.

Case Study


Medical Procedure Certification

P
-
15

© 2005 NeuralWare. All rights reserved.

Case Study


Medical Procedure Certification

Acquire/Validate Case Input

Retrieve Metrics

Select/Execute Model

Apply Thresholds

Approve Procedure?

Process Manually

Update Metrics

Selected for Audit?

NO

YES

YES

DONE

Metric Database

P
-
16

© 2005 NeuralWare. All rights reserved.

Case Study


Crime Forecasting

Objectives


Identify Patterns in Criminal Activity that indicate Potential Future Trouble Spots


Redirect Police Resources to Focus on Areas where Serious Crime is expected to Increase

Challenges


Defining Crime Categories and Severity Levels


Inconsistent data formats and labeling; missing or non
-
existent data

Process


Used NeuralSight to Build and Evaluate ~ 10,000 Models in 1 week


On
-
going evaluation by researchers at Carnegie Mellon University


P
-
17

© 2005 NeuralWare. All rights reserved.

Case Study


Crime Forecasting



P
-
18

© 2005 NeuralWare. All rights reserved.

Case Study


Crime Forecasting

How to Forecast Change in Crime


Police know current crime levels


Have allocated resources to respond to existing crimes

Most valuable information for tactical level planning:


Where is crime likely to have large increases next month?


Forecast crime by area and calculate:


Forecasted Change (t+1) = Forecast (t+1)


Actual (t)



The Benefit



Better Allocation of Scarce Resources



P
-
19

© 2005 NeuralWare. All rights reserved.

Case Study


Crime Forecasting

Forecasted Change for July

P
-
20

© 2005 NeuralWare. All rights reserved.

Case Study


Grain Quality Assessment

Objectives


Provide a Platform for rapidly and consistently assessing the quality of grain


Maintain detailed records of tests and build foundation for data mining

Challenges


No “Gold Standard” for decisions


even experienced human inspectors are inconsistent


Requires tedious work to identify wide variety of training data samples

Process


Used Predict and NeuralSight to Build and Evaluate many thousands of Models


Now developing image database to support agriculture research


P
-
21

© 2005 NeuralWare. All rights reserved.

Case Study


Grain Quality Assessment

An Instrument


and examples of seed images

P
-
22

© 2005 NeuralWare. All rights reserved.

Case Study


Grain Quality Assessment


Many (more than 300) initial features per seed



Predict Variable Selection found a much smaller set of

features to use in building models


The characteristics of grain that are important are
difficult even for human inspectors to identify



Multiple neural networks are used to make the hard decisions


The value of wheat and other commodities depends on
its quality


millions of dollars are at risk if quality
decisions are incorrect!


P
-
23

© 2005 NeuralWare. All rights reserved.

What Have you Learned?


Neural Networks make Powerful Decision Support Systems


Human Judgment Determines the Cost/Benefit Tradeoff for Accuracy


Know your Problem !


Neural Network Decisions are based on Learning Patterns



Relationships in Historical Data are the basis for Current Action


Know your Data !



P
-
24

© 2005 NeuralWare. All rights reserved.

Thank You !










Jack Copper

NeuralWare

jack.copper@neuralware.com


Hiroshi Maruyama

SET Software Co. Ltd.


mal@setsw.co.jp