Transcript Predstavitev Laboratorija za Kognitivno modeliranje

Artificial Intelligence and
Cognitive Modeling
Laboratory for Cognitive Modeling
4.11.2011
lkm.fri.uni-lj.si
Terminology, terminology…
Artificial Intelligence
Machine Learning
Data Mining
Cognitive Modeling
lkm.fri.uni-lj.si
Data modeling
Different types of data
from different sources
Data mining
Data model
Credit ranking (1=default)
Cat.
%
Bad 52.01
Good 47.99
Total (100.00)
n
168
155
323
Paid Weekly/Monthly
P-value=0.0000, Chi-square=179.6665, df=1
Weekly pay
Monthly salary
Cat.
%
n
Bad 86.67 143
Good 13.33 22
Total (51.08) 165
Cat.
%
n
Bad 15.82 25
Good 84.18 133
Total (48.92) 158
Age Categorical
P-value=0.0000, Chi-square=30.1113, df=1
Young (< 25);Middle (25-35)
Cat.
%
n
Bad 90.51 143
Good 9.49
15
Total (48.92) 158
Age Categorical
P-value=0.0000, Chi-square=58.7255, df=1
Old ( > 35)
Cat.
%
Bad
0.00
Good 100.00
Total (2.17)
n
0
7
7
Young (< 25)
Middle (25-35);Old ( > 35)
Cat.
%
n
Bad 48.98 24
Good 51.02 25
Total (15.17) 49
Cat.
%
n
Bad
0.92
1
Good 99.08 108
Total (33.75) 109
Social Class
P-value=0.0016, Chi-square=12.0388, df=1
Management;Clerical
Cat.
%
Bad
0.00
Good 100.00
Total (2.48)
n
0
8
8
Professional
Cat.
%
n
Bad 58.54 24
Good 41.46 17
Total (12.69) 41
Background knowledge
lkm.fri.uni-lj.si
Models and their use
• Supervised and
unsupervised modeling
• Model types:
– decision trees and
decision rules
– artificial neural networks
– regression trees
– nearest neighbors
– association rules
– random forests
– …
• Different models,
different use:
– model structure
(presentation of the
relationship between
inputs and outputs)
– prediction
– associations (relationships)
between input values
– clustering
– outlier detection
– …
lkm.fri.uni-lj.si
Example:
applications in medical
diagnostics and prognostics
• modeling the knowledge
and skills of specialist
physicians
• using models for
decision support
• scintigraphy of the
skeleton and heart,
oncology, traumatology, …
lkm.fri.uni-lj.si
Medical diagnostics and prognostics
• Input: background knowledge, descriptions of patients
with subsequently confirmed diagnosis
• How to diagnose?
• How to predict the occurrence of a disease
or its recurrence?
• Very good results in specialized areas
(significantly better than specialists).
• What characteristics have the greatest impact
on the disease?
• What is the reliability of computer predictions
(diagnosis and prognosis)?
• How to explain predictions and bring them
closer to doctors?
lkm.fri.uni-lj.si
Reliability estimation for medical
diagnosis
General
methods for
estimating the
reliability of
individual
predictions are
developed.
lkm.fri.uni-lj.si
Diagnosis explanation
General methods
for explaining
predictions are
developed.
lkm.fri.uni-lj.si
Skeletal pathology detection
• Skeletal scintigraphy
• Background knowledge
of human anatomy
• Known diagnoses
lkm.fri.uni-lj.si
Diagnosis of coronary artery disease
• Heart scintigraphy
• Input data in the
form of images
• Medical records
• Reliability estimates
lkm.fri.uni-lj.si
Marketing
• How do customers decide
what products to buy?
• How to arrange ads in an
optimal way?
• When is the best time to
broadcast television ads?
lkm.fri.uni-lj.si
Advanced sports analysis
• A basketball match simulation
• in collaboration with the
Faculty of Sport in Ljubljana :
– analysis of the impact of rules
changes in 2010/11 season
lkm.fri.uni-lj.si
And more...
• prediction market
• clickstream analysis
• façade analysis
−1.0
0.0
1.0
• prediction intervals
400
600
800
1000
0.0 0.5 1.0
200
−1.0
0
0
200
400
600
800
1000
lkm.fri.uni-lj.si
Conclusion
• Versatile applicability of artificial intelligence
methods, especially data mining
– ability to process large amounts of data
– variety of data types
– inclusion of background knowledge
• However:
Artificial Intelligence (still) is not intelligence
lkm.fri.uni-lj.si
Scientific and developmental
competence
We are the authors of numerous papers
in scientific journals and books
(over 700 citations)
We regularly participate
at scientific conferences
and present our work
We are members of editorial boards
and program committees
We have a long experience
in the field of medicine, marketing,
financial sector, telecommunications ...
lkm.fri.uni-lj.si
Collaboration with other institutions
University of Hasselt
Institute of Oncology AD Consulting Bion Institute
Jožef Stefan Institute-department of knowledge technologies
Starcom The Laboratory of Neuroendocrinology
Clinic for Nuclear Medicine Intensio Faculty of sports
ASCR Institute of Computer Science
University of Porto
University of Kragujevac
University of Ioannina
University of Malaga
lkm.fri.uni-lj.si
Who are we?
Darko Pevec
doc. dr. Zoran Bosnić
izr. prof. dr. Marko Robnik Šikonja
doc. dr. Matjaž Kukar
prof. dr. Igor Kononenko
Domen Košir
dr. Erik Štrumbelj
Miha Drole
as. mag. Petar Vračar
as. Matej Pičulin
lkm.fri.uni-lj.si