SlideShare a Scribd company logo
Towards Evidence Terminological Decision Trees
15th International Conference on Information Processing and
Management of Uncertainty in Knowledge-Based Systems
Giuseppe Rizzo, Claudia d’Amato, Nicola Fanizzi and Floriana Esposito
Dipartimento di Informatica
Universit`a degli Studi di Bari ”Aldo Moro”, Bari, Italy
July 15 - 19, 2014
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 1 / 17
Outline
1 Introduction & Motivation
2 The approach
3 Evaluation
4 Conclusions
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 2 / 17
Introduction & Motivation
Introduction
In the context of Web of Data, machine learning algorithms can
support:
the ontology completion
the development of new non-standard inference services
by exploiting regularities in the a knowledge base
Lack of disjointness axioms in ontologies
The Open World Assumption does not allows to assess the membership
w.r.t a query concept (or its complement) deductively
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 3 / 17
Introduction & Motivation
Introduction
Some techniques proposed in literature are inspired from Inductive
Logic Program
E.g.: Terminological Decision Trees (TDT) induction algorithm
In this kind of methods the uncertainty is not considered explicitly
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 4 / 17
Introduction & Motivation
Terminological Decision Trees
Let K = (T , A), a Terminological Decision Tree (TDT) is a binary tree
where:
each node contains a conjunctive concept description D;
each departing edge is the result of a class-membership test w.r.t. D,
i.e., given an individual a, K |= D(a)?
if a node with E is the father of the node with D then D is obtained
by using a refinement operator and one of the following conditions
should be verified:
D introduces a new concept name,
D is an existential restriction,
D is an universal restriction of any its ancestor.
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 5 / 17
Introduction & Motivation
Motivations
Given the problem to predict the membership w.r.t. either the
concept or its complement, TDTs cannot to determine the result
w.r.t. the intermediate test.
due to the treatment of missing values with DTs
What happens when neither the membership for ∃hasPart. nor for the
complement can be decided?
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 6 / 17
The approach
The approach
Extending Terminological Decision Trees with the Dempster-Shafer
Theory the problem can be overcome
Dempster-Shafer Theory (DST) is a more suitable framework than
the probabilistic one because it allows to represents explicitly the
ignorance related to the Open World Assumption
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 7 / 17
The approach
Semantic Web Knowledge bases
Semantic Web knowledge bases can be modeled by using specific
fragments of Description Logics (DL)
A domain is modeled through primitive concepts (classes) and roles
(relations),
A knowledge base is a couple K = (T , A) where
T (TBox) contains axioms concerning concepts and roles
A (ABox) contains factual knowledge (C(a), resp. R(a, b)).
The set of individuals occurring in A is denoted by Ind(A)
Various reasoning services are available
instance checking: a service to decide if an individual is an instance of
a concept or not
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 8 / 17
The approach
The problem
Given:
a knowledge base K = (T , A)
a target concept C,
the sets of positive and negative examples for C:
Ps = {a ∈ Ind(A) | K |= C(a)} and Ns = {a ∈ Ind(A) | K |= ¬C(a)}
Obtain:
a concept description D for C (C D), such that:
K |= D(a) ∀a ∈ Ps
K |= ¬D(a) ∀a ∈ Ns
The intensional definition should be general enough to predict the
membership for future instances.
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 9 / 17
The approach
DST-Terminological Decision Trees
Let K = (T , A), a DST-Terminological Decision Tree (DST-TDT) is a
binary tree where:
each node contains a conjunctive concept description D and a Basic
Belief Assignement (BBA) m obtained by counting the positive,
negative and uncertain instances;
each departing edge is the result of a class-membership test w.r.t. D,
i.e., given an individual a, K |= D(a)?
if a node having the concept description E is the father of the node
with the concept description D then D is obtained by using the same
operator for Terminological Decision Tree
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 10 / 17
The approach
DST- Terminological Decision Tree
Training
Starting from the root the method refines the concept description
installed into the current node
Various pairs (D, m) are generated (m is generated from the frame of
discernement Ω = {D, ¬D} by counting positive, negative and
uncertain-membership instances that reached the node
Best Concept: the one having the most definite membership (i.e. the
smaller number of uncertain-membership instances)
non − specificity(D) =
A⊆ΩD
m(A)log|A|
Split the instances according to the results of the instance check test
Stop conditions:
the node is pure w.r.t. the membership
Non-specificity measure goes beyond a thresold ν
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 11 / 17
The approach
DST - Terminological Decision Tree
How-to use DST-TDTs
The membership prediction for a new indivdual is done by exploring
the tree according to the instance check test K |= C(a) and
K |= ¬C(a)
if neither the first test nor the second test return a positive answer
(due to OWA) both branches are followed
all the BBAs associated to the reached leaves are pooled according to a
DST combination rule
the class is the one which maximizes the Confirmation Function
∀A ⊆ Ω Conf (A) = Bel(A) + Pl(A) − 1
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 12 / 17
Evaluation
Evaluation
Setting
30 randomly generated queries
0.632 bootstrap
growth control threshold ν = 0.5
pooling by means of Dubois-Prade Combination Rule
Using a reasoner to decide the ground truth:
match: rate of the test cases (individuals) for which the inductive
model and a reasoner predict the same membership (i.e. +1 | +1,
−1 | −1, 0 | 0);
commission: rate of the cases for which predictions are opposite (i.e.
+1 | −1, −1 | +1);
omission: rate of test cases for which the inductive method cannot
determine a definite membership (−1, +1) while the reasoner is able to
do it;
induction: rate of cases where the inductive method can predict a
membership while it is not logically derivable.
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 13 / 17
Evaluation
Evaluation
Results
Comparison between the original terminological trees (DLTree) and the
DST-TDTs both by with the threshold (DSTG) and without (DSTTree)
Ontology Index DLTree DSTTree DSTG
FSM
match 95.34±04.94 93.22±07.33 86.16±10.48
commiss. 01.81±02.18 01.67±03.05 02.07±03.19
omission 00.74±02.15 02.57±04.09 04.98±05.99
induction 02.11±04.42 02.54±01.89 01.16±01.26
Leo
match 95.53±10.07 97.07±04.55 94.61±06.75
commiss. 00.48±00.57 00.41±00.86 00.41±01.00
omission 03.42±09.84 01.94±04.38 00.58±00.51
induction 00.57±03.13 00.58±00.51 00.00±00.00
LUBM
match 20.78±00.11 79.23±00.11 79.22±00.12
commiss. 00.00±00.00 00.00±00.00 00.00±00.00
omission 00.00±00.00 20.77±00.11 20.78±00.12
induction 79.22±00.11 00.00±00.00 00.00±00.00
BioPax
match 96.87±07.35 85.76±21.60 82.15±21.10
commiss. 01.63±06.44 11.81±19.96 12.32±19.90
omission 00.30±00.98 01.54±03.02 04.88±03.03
induction 01.21±00.56 00.89±00.53 00.26±00.27
NTN
match 27.02±01.91 18.97±19.01 87.63±00.19
commiss. 00.00±00.00 00.39±01.08 00.00±00.00
omission 00.22±00.26 02.09±03.00 12.37±00.19
induction 72.77±01.51 78.54±17.34 00.00±00.00
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 14 / 17
Evaluation
Discussion
Preliminary evaluation showed that the method did not overcome the
original version
Large number of omission cases: a more conservative behavior
Probably, due to the combination rules that returned a pooled BBA
with a greater value in favor of ignorance
the threshold employed to control the growth reduced the induction
cases
Results are not stable yet
large standard deviation
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 15 / 17
Conclusions
Conclusion & Extensions
An extension of TDT based on Dempster-Shafer Theory has been
proposed and a preliminary evaluation has been made.
The results are not very good, yet
Extensions:
Tackle the compexity of the model
Pruning algorithms
Consider other measures for the selection of the best candidates
Further datasets should be considered
Linked Data datasets
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 16 / 17
Conclusions
Thank you!
Questions?
G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 17 / 17

More Related Content

Similar to Towards Evidence Terminological Decision Tree

Strategies for Cooperation Emergence in Distributed Service Discovery
Strategies for Cooperation Emergence in Distributed Service DiscoveryStrategies for Cooperation Emergence in Distributed Service Discovery
Strategies for Cooperation Emergence in Distributed Service Discovery
Miguel Rebollo
 
Star Scholars_Poster
Star Scholars_PosterStar Scholars_Poster
Star Scholars_Poster
Katya Hristova
 
On cascading small decision trees
On cascading small decision treesOn cascading small decision trees
On cascading small decision trees
Julià Minguillón
 
A Data-driven Method for the Detection of Close Submitters in Online Learning...
A Data-driven Method for the Detection of Close Submitters in Online Learning...A Data-driven Method for the Detection of Close Submitters in Online Learning...
A Data-driven Method for the Detection of Close Submitters in Online Learning...
MIT
 
LDA on social bookmarking systems
LDA on social bookmarking systemsLDA on social bookmarking systems
LDA on social bookmarking systems
Denis Parra Santander
 
Bayesian/Fiducial/Frequentist Uncertainty Quantification by Artificial Samples
Bayesian/Fiducial/Frequentist Uncertainty Quantification by Artificial SamplesBayesian/Fiducial/Frequentist Uncertainty Quantification by Artificial Samples
Bayesian/Fiducial/Frequentist Uncertainty Quantification by Artificial Samples
Liverpool Institute for Risk and Uncertainty
 
Engineering Data Science Objectives for Social Network Analysis
Engineering Data Science Objectives for Social Network AnalysisEngineering Data Science Objectives for Social Network Analysis
Engineering Data Science Objectives for Social Network Analysis
David Gleich
 
Algoritma Random Forest beserta aplikasi nya
Algoritma Random Forest beserta aplikasi nyaAlgoritma Random Forest beserta aplikasi nya
Algoritma Random Forest beserta aplikasi nya
batubao
 
6.04218.062J Mathematics for Computer Science September 9, 20.docx
6.04218.062J Mathematics for Computer Science September 9, 20.docx6.04218.062J Mathematics for Computer Science September 9, 20.docx
6.04218.062J Mathematics for Computer Science September 9, 20.docx
troutmanboris
 
Approximating Numeric Role Fillers via Predictive Clustering Trees for Know...
Approximating Numeric Role Fillers via Predictive Clustering Trees  for  Know...Approximating Numeric Role Fillers via Predictive Clustering Trees  for  Know...
Approximating Numeric Role Fillers via Predictive Clustering Trees for Know...
Giuseppe Rizzo
 
Finding and Quantifying Temporal-Aware Contradiction in Reviews
Finding and Quantifying Temporal-Aware Contradiction in ReviewsFinding and Quantifying Temporal-Aware Contradiction in Reviews
Finding and Quantifying Temporal-Aware Contradiction in Reviews
Ismail BADACHE
 
Harnessing Ratings and Aspect-Sentiment to Estimate Contradiction Intensity i...
Harnessing Ratings and Aspect-Sentiment to Estimate Contradiction Intensity i...Harnessing Ratings and Aspect-Sentiment to Estimate Contradiction Intensity i...
Harnessing Ratings and Aspect-Sentiment to Estimate Contradiction Intensity i...
Ismail BADACHE
 
11.selection method by fuzzy set theory and preference matrix
11.selection method by fuzzy set theory and preference matrix11.selection method by fuzzy set theory and preference matrix
11.selection method by fuzzy set theory and preference matrix
Alexander Decker
 
Selection method by fuzzy set theory and preference matrix
Selection method by fuzzy set theory and preference matrixSelection method by fuzzy set theory and preference matrix
Selection method by fuzzy set theory and preference matrix
Alexander Decker
 
Cluster
ClusterCluster
DEseq, voom and vst
DEseq, voom and vstDEseq, voom and vst
DEseq, voom and vst
Qiang Kou
 
1_Introduction_printable.pdf
1_Introduction_printable.pdf1_Introduction_printable.pdf
1_Introduction_printable.pdf
Elio Laureano
 
Social Learning in Networks: Extraction Deterministic Rules
Social Learning in Networks: Extraction Deterministic RulesSocial Learning in Networks: Extraction Deterministic Rules
Social Learning in Networks: Extraction Deterministic Rules
Dmitrii Ignatov
 
Problem solving content
Problem solving contentProblem solving content
Problem solving content
Timothy Welsh
 
problem_solving in physics
 problem_solving in physics problem_solving in physics
problem_solving in physics
Timothy Welsh
 

Similar to Towards Evidence Terminological Decision Tree (20)

Strategies for Cooperation Emergence in Distributed Service Discovery
Strategies for Cooperation Emergence in Distributed Service DiscoveryStrategies for Cooperation Emergence in Distributed Service Discovery
Strategies for Cooperation Emergence in Distributed Service Discovery
 
Star Scholars_Poster
Star Scholars_PosterStar Scholars_Poster
Star Scholars_Poster
 
On cascading small decision trees
On cascading small decision treesOn cascading small decision trees
On cascading small decision trees
 
A Data-driven Method for the Detection of Close Submitters in Online Learning...
A Data-driven Method for the Detection of Close Submitters in Online Learning...A Data-driven Method for the Detection of Close Submitters in Online Learning...
A Data-driven Method for the Detection of Close Submitters in Online Learning...
 
LDA on social bookmarking systems
LDA on social bookmarking systemsLDA on social bookmarking systems
LDA on social bookmarking systems
 
Bayesian/Fiducial/Frequentist Uncertainty Quantification by Artificial Samples
Bayesian/Fiducial/Frequentist Uncertainty Quantification by Artificial SamplesBayesian/Fiducial/Frequentist Uncertainty Quantification by Artificial Samples
Bayesian/Fiducial/Frequentist Uncertainty Quantification by Artificial Samples
 
Engineering Data Science Objectives for Social Network Analysis
Engineering Data Science Objectives for Social Network AnalysisEngineering Data Science Objectives for Social Network Analysis
Engineering Data Science Objectives for Social Network Analysis
 
Algoritma Random Forest beserta aplikasi nya
Algoritma Random Forest beserta aplikasi nyaAlgoritma Random Forest beserta aplikasi nya
Algoritma Random Forest beserta aplikasi nya
 
6.04218.062J Mathematics for Computer Science September 9, 20.docx
6.04218.062J Mathematics for Computer Science September 9, 20.docx6.04218.062J Mathematics for Computer Science September 9, 20.docx
6.04218.062J Mathematics for Computer Science September 9, 20.docx
 
Approximating Numeric Role Fillers via Predictive Clustering Trees for Know...
Approximating Numeric Role Fillers via Predictive Clustering Trees  for  Know...Approximating Numeric Role Fillers via Predictive Clustering Trees  for  Know...
Approximating Numeric Role Fillers via Predictive Clustering Trees for Know...
 
Finding and Quantifying Temporal-Aware Contradiction in Reviews
Finding and Quantifying Temporal-Aware Contradiction in ReviewsFinding and Quantifying Temporal-Aware Contradiction in Reviews
Finding and Quantifying Temporal-Aware Contradiction in Reviews
 
Harnessing Ratings and Aspect-Sentiment to Estimate Contradiction Intensity i...
Harnessing Ratings and Aspect-Sentiment to Estimate Contradiction Intensity i...Harnessing Ratings and Aspect-Sentiment to Estimate Contradiction Intensity i...
Harnessing Ratings and Aspect-Sentiment to Estimate Contradiction Intensity i...
 
11.selection method by fuzzy set theory and preference matrix
11.selection method by fuzzy set theory and preference matrix11.selection method by fuzzy set theory and preference matrix
11.selection method by fuzzy set theory and preference matrix
 
Selection method by fuzzy set theory and preference matrix
Selection method by fuzzy set theory and preference matrixSelection method by fuzzy set theory and preference matrix
Selection method by fuzzy set theory and preference matrix
 
Cluster
ClusterCluster
Cluster
 
DEseq, voom and vst
DEseq, voom and vstDEseq, voom and vst
DEseq, voom and vst
 
1_Introduction_printable.pdf
1_Introduction_printable.pdf1_Introduction_printable.pdf
1_Introduction_printable.pdf
 
Social Learning in Networks: Extraction Deterministic Rules
Social Learning in Networks: Extraction Deterministic RulesSocial Learning in Networks: Extraction Deterministic Rules
Social Learning in Networks: Extraction Deterministic Rules
 
Problem solving content
Problem solving contentProblem solving content
Problem solving content
 
problem_solving in physics
 problem_solving in physics problem_solving in physics
problem_solving in physics
 

Recently uploaded

DATA COMMS-NETWORKS YR2 lecture 08 NAT & CLOUD.docx
DATA COMMS-NETWORKS YR2 lecture 08 NAT & CLOUD.docxDATA COMMS-NETWORKS YR2 lecture 08 NAT & CLOUD.docx
DATA COMMS-NETWORKS YR2 lecture 08 NAT & CLOUD.docx
SaffaIbrahim1
 
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdfUdemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
Fernanda Palhano
 
Build applications with generative AI on Google Cloud
Build applications with generative AI on Google CloudBuild applications with generative AI on Google Cloud
Build applications with generative AI on Google Cloud
Márton Kodok
 
End-to-end pipeline agility - Berlin Buzzwords 2024
End-to-end pipeline agility - Berlin Buzzwords 2024End-to-end pipeline agility - Berlin Buzzwords 2024
End-to-end pipeline agility - Berlin Buzzwords 2024
Lars Albertsson
 
"Financial Odyssey: Navigating Past Performance Through Diverse Analytical Lens"
"Financial Odyssey: Navigating Past Performance Through Diverse Analytical Lens""Financial Odyssey: Navigating Past Performance Through Diverse Analytical Lens"
"Financial Odyssey: Navigating Past Performance Through Diverse Analytical Lens"
sameer shah
 
A presentation that explain the Power BI Licensing
A presentation that explain the Power BI LicensingA presentation that explain the Power BI Licensing
A presentation that explain the Power BI Licensing
AlessioFois2
 
一比一原版(UO毕业证)渥太华大学毕业证如何办理
一比一原版(UO毕业证)渥太华大学毕业证如何办理一比一原版(UO毕业证)渥太华大学毕业证如何办理
一比一原版(UO毕业证)渥太华大学毕业证如何办理
aqzctr7x
 
Learn SQL from basic queries to Advance queries
Learn SQL from basic queries to Advance queriesLearn SQL from basic queries to Advance queries
Learn SQL from basic queries to Advance queries
manishkhaire30
 
STATATHON: Unleashing the Power of Statistics in a 48-Hour Knowledge Extravag...
STATATHON: Unleashing the Power of Statistics in a 48-Hour Knowledge Extravag...STATATHON: Unleashing the Power of Statistics in a 48-Hour Knowledge Extravag...
STATATHON: Unleashing the Power of Statistics in a 48-Hour Knowledge Extravag...
sameer shah
 
writing report business partner b1+ .pdf
writing report business partner b1+ .pdfwriting report business partner b1+ .pdf
writing report business partner b1+ .pdf
VyNguyen709676
 
Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging DataPredictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
Kiwi Creative
 
Analysis insight about a Flyball dog competition team's performance
Analysis insight about a Flyball dog competition team's performanceAnalysis insight about a Flyball dog competition team's performance
Analysis insight about a Flyball dog competition team's performance
roli9797
 
University of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma TranscriptUniversity of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma Transcript
soxrziqu
 
原版制作(unimelb毕业证书)墨尔本大学毕业证Offer一模一样
原版制作(unimelb毕业证书)墨尔本大学毕业证Offer一模一样原版制作(unimelb毕业证书)墨尔本大学毕业证Offer一模一样
原版制作(unimelb毕业证书)墨尔本大学毕业证Offer一模一样
ihavuls
 
一比一原版兰加拉学院毕业证(Langara毕业证书)学历如何办理
一比一原版兰加拉学院毕业证(Langara毕业证书)学历如何办理一比一原版兰加拉学院毕业证(Langara毕业证书)学历如何办理
一比一原版兰加拉学院毕业证(Langara毕业证书)学历如何办理
hyfjgavov
 
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
nuttdpt
 
06-12-2024-BudapestDataForum-BuildingReal-timePipelineswithFLaNK AIM
06-12-2024-BudapestDataForum-BuildingReal-timePipelineswithFLaNK AIM06-12-2024-BudapestDataForum-BuildingReal-timePipelineswithFLaNK AIM
06-12-2024-BudapestDataForum-BuildingReal-timePipelineswithFLaNK AIM
Timothy Spann
 
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
v7oacc3l
 
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
mkkikqvo
 
The Ipsos - AI - Monitor 2024 Report.pdf
The  Ipsos - AI - Monitor 2024 Report.pdfThe  Ipsos - AI - Monitor 2024 Report.pdf
The Ipsos - AI - Monitor 2024 Report.pdf
Social Samosa
 

Recently uploaded (20)

DATA COMMS-NETWORKS YR2 lecture 08 NAT & CLOUD.docx
DATA COMMS-NETWORKS YR2 lecture 08 NAT & CLOUD.docxDATA COMMS-NETWORKS YR2 lecture 08 NAT & CLOUD.docx
DATA COMMS-NETWORKS YR2 lecture 08 NAT & CLOUD.docx
 
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdfUdemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
 
Build applications with generative AI on Google Cloud
Build applications with generative AI on Google CloudBuild applications with generative AI on Google Cloud
Build applications with generative AI on Google Cloud
 
End-to-end pipeline agility - Berlin Buzzwords 2024
End-to-end pipeline agility - Berlin Buzzwords 2024End-to-end pipeline agility - Berlin Buzzwords 2024
End-to-end pipeline agility - Berlin Buzzwords 2024
 
"Financial Odyssey: Navigating Past Performance Through Diverse Analytical Lens"
"Financial Odyssey: Navigating Past Performance Through Diverse Analytical Lens""Financial Odyssey: Navigating Past Performance Through Diverse Analytical Lens"
"Financial Odyssey: Navigating Past Performance Through Diverse Analytical Lens"
 
A presentation that explain the Power BI Licensing
A presentation that explain the Power BI LicensingA presentation that explain the Power BI Licensing
A presentation that explain the Power BI Licensing
 
一比一原版(UO毕业证)渥太华大学毕业证如何办理
一比一原版(UO毕业证)渥太华大学毕业证如何办理一比一原版(UO毕业证)渥太华大学毕业证如何办理
一比一原版(UO毕业证)渥太华大学毕业证如何办理
 
Learn SQL from basic queries to Advance queries
Learn SQL from basic queries to Advance queriesLearn SQL from basic queries to Advance queries
Learn SQL from basic queries to Advance queries
 
STATATHON: Unleashing the Power of Statistics in a 48-Hour Knowledge Extravag...
STATATHON: Unleashing the Power of Statistics in a 48-Hour Knowledge Extravag...STATATHON: Unleashing the Power of Statistics in a 48-Hour Knowledge Extravag...
STATATHON: Unleashing the Power of Statistics in a 48-Hour Knowledge Extravag...
 
writing report business partner b1+ .pdf
writing report business partner b1+ .pdfwriting report business partner b1+ .pdf
writing report business partner b1+ .pdf
 
Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging DataPredictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
 
Analysis insight about a Flyball dog competition team's performance
Analysis insight about a Flyball dog competition team's performanceAnalysis insight about a Flyball dog competition team's performance
Analysis insight about a Flyball dog competition team's performance
 
University of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma TranscriptUniversity of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma Transcript
 
原版制作(unimelb毕业证书)墨尔本大学毕业证Offer一模一样
原版制作(unimelb毕业证书)墨尔本大学毕业证Offer一模一样原版制作(unimelb毕业证书)墨尔本大学毕业证Offer一模一样
原版制作(unimelb毕业证书)墨尔本大学毕业证Offer一模一样
 
一比一原版兰加拉学院毕业证(Langara毕业证书)学历如何办理
一比一原版兰加拉学院毕业证(Langara毕业证书)学历如何办理一比一原版兰加拉学院毕业证(Langara毕业证书)学历如何办理
一比一原版兰加拉学院毕业证(Langara毕业证书)学历如何办理
 
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
 
06-12-2024-BudapestDataForum-BuildingReal-timePipelineswithFLaNK AIM
06-12-2024-BudapestDataForum-BuildingReal-timePipelineswithFLaNK AIM06-12-2024-BudapestDataForum-BuildingReal-timePipelineswithFLaNK AIM
06-12-2024-BudapestDataForum-BuildingReal-timePipelineswithFLaNK AIM
 
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
 
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
 
The Ipsos - AI - Monitor 2024 Report.pdf
The  Ipsos - AI - Monitor 2024 Report.pdfThe  Ipsos - AI - Monitor 2024 Report.pdf
The Ipsos - AI - Monitor 2024 Report.pdf
 

Towards Evidence Terminological Decision Tree

  • 1. Towards Evidence Terminological Decision Trees 15th International Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems Giuseppe Rizzo, Claudia d’Amato, Nicola Fanizzi and Floriana Esposito Dipartimento di Informatica Universit`a degli Studi di Bari ”Aldo Moro”, Bari, Italy July 15 - 19, 2014 G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 1 / 17
  • 2. Outline 1 Introduction & Motivation 2 The approach 3 Evaluation 4 Conclusions G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 2 / 17
  • 3. Introduction & Motivation Introduction In the context of Web of Data, machine learning algorithms can support: the ontology completion the development of new non-standard inference services by exploiting regularities in the a knowledge base Lack of disjointness axioms in ontologies The Open World Assumption does not allows to assess the membership w.r.t a query concept (or its complement) deductively G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 3 / 17
  • 4. Introduction & Motivation Introduction Some techniques proposed in literature are inspired from Inductive Logic Program E.g.: Terminological Decision Trees (TDT) induction algorithm In this kind of methods the uncertainty is not considered explicitly G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 4 / 17
  • 5. Introduction & Motivation Terminological Decision Trees Let K = (T , A), a Terminological Decision Tree (TDT) is a binary tree where: each node contains a conjunctive concept description D; each departing edge is the result of a class-membership test w.r.t. D, i.e., given an individual a, K |= D(a)? if a node with E is the father of the node with D then D is obtained by using a refinement operator and one of the following conditions should be verified: D introduces a new concept name, D is an existential restriction, D is an universal restriction of any its ancestor. G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 5 / 17
  • 6. Introduction & Motivation Motivations Given the problem to predict the membership w.r.t. either the concept or its complement, TDTs cannot to determine the result w.r.t. the intermediate test. due to the treatment of missing values with DTs What happens when neither the membership for ∃hasPart. nor for the complement can be decided? G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 6 / 17
  • 7. The approach The approach Extending Terminological Decision Trees with the Dempster-Shafer Theory the problem can be overcome Dempster-Shafer Theory (DST) is a more suitable framework than the probabilistic one because it allows to represents explicitly the ignorance related to the Open World Assumption G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 7 / 17
  • 8. The approach Semantic Web Knowledge bases Semantic Web knowledge bases can be modeled by using specific fragments of Description Logics (DL) A domain is modeled through primitive concepts (classes) and roles (relations), A knowledge base is a couple K = (T , A) where T (TBox) contains axioms concerning concepts and roles A (ABox) contains factual knowledge (C(a), resp. R(a, b)). The set of individuals occurring in A is denoted by Ind(A) Various reasoning services are available instance checking: a service to decide if an individual is an instance of a concept or not G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 8 / 17
  • 9. The approach The problem Given: a knowledge base K = (T , A) a target concept C, the sets of positive and negative examples for C: Ps = {a ∈ Ind(A) | K |= C(a)} and Ns = {a ∈ Ind(A) | K |= ¬C(a)} Obtain: a concept description D for C (C D), such that: K |= D(a) ∀a ∈ Ps K |= ¬D(a) ∀a ∈ Ns The intensional definition should be general enough to predict the membership for future instances. G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 9 / 17
  • 10. The approach DST-Terminological Decision Trees Let K = (T , A), a DST-Terminological Decision Tree (DST-TDT) is a binary tree where: each node contains a conjunctive concept description D and a Basic Belief Assignement (BBA) m obtained by counting the positive, negative and uncertain instances; each departing edge is the result of a class-membership test w.r.t. D, i.e., given an individual a, K |= D(a)? if a node having the concept description E is the father of the node with the concept description D then D is obtained by using the same operator for Terminological Decision Tree G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 10 / 17
  • 11. The approach DST- Terminological Decision Tree Training Starting from the root the method refines the concept description installed into the current node Various pairs (D, m) are generated (m is generated from the frame of discernement Ω = {D, ¬D} by counting positive, negative and uncertain-membership instances that reached the node Best Concept: the one having the most definite membership (i.e. the smaller number of uncertain-membership instances) non − specificity(D) = A⊆ΩD m(A)log|A| Split the instances according to the results of the instance check test Stop conditions: the node is pure w.r.t. the membership Non-specificity measure goes beyond a thresold ν G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 11 / 17
  • 12. The approach DST - Terminological Decision Tree How-to use DST-TDTs The membership prediction for a new indivdual is done by exploring the tree according to the instance check test K |= C(a) and K |= ¬C(a) if neither the first test nor the second test return a positive answer (due to OWA) both branches are followed all the BBAs associated to the reached leaves are pooled according to a DST combination rule the class is the one which maximizes the Confirmation Function ∀A ⊆ Ω Conf (A) = Bel(A) + Pl(A) − 1 G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 12 / 17
  • 13. Evaluation Evaluation Setting 30 randomly generated queries 0.632 bootstrap growth control threshold ν = 0.5 pooling by means of Dubois-Prade Combination Rule Using a reasoner to decide the ground truth: match: rate of the test cases (individuals) for which the inductive model and a reasoner predict the same membership (i.e. +1 | +1, −1 | −1, 0 | 0); commission: rate of the cases for which predictions are opposite (i.e. +1 | −1, −1 | +1); omission: rate of test cases for which the inductive method cannot determine a definite membership (−1, +1) while the reasoner is able to do it; induction: rate of cases where the inductive method can predict a membership while it is not logically derivable. G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 13 / 17
  • 14. Evaluation Evaluation Results Comparison between the original terminological trees (DLTree) and the DST-TDTs both by with the threshold (DSTG) and without (DSTTree) Ontology Index DLTree DSTTree DSTG FSM match 95.34±04.94 93.22±07.33 86.16±10.48 commiss. 01.81±02.18 01.67±03.05 02.07±03.19 omission 00.74±02.15 02.57±04.09 04.98±05.99 induction 02.11±04.42 02.54±01.89 01.16±01.26 Leo match 95.53±10.07 97.07±04.55 94.61±06.75 commiss. 00.48±00.57 00.41±00.86 00.41±01.00 omission 03.42±09.84 01.94±04.38 00.58±00.51 induction 00.57±03.13 00.58±00.51 00.00±00.00 LUBM match 20.78±00.11 79.23±00.11 79.22±00.12 commiss. 00.00±00.00 00.00±00.00 00.00±00.00 omission 00.00±00.00 20.77±00.11 20.78±00.12 induction 79.22±00.11 00.00±00.00 00.00±00.00 BioPax match 96.87±07.35 85.76±21.60 82.15±21.10 commiss. 01.63±06.44 11.81±19.96 12.32±19.90 omission 00.30±00.98 01.54±03.02 04.88±03.03 induction 01.21±00.56 00.89±00.53 00.26±00.27 NTN match 27.02±01.91 18.97±19.01 87.63±00.19 commiss. 00.00±00.00 00.39±01.08 00.00±00.00 omission 00.22±00.26 02.09±03.00 12.37±00.19 induction 72.77±01.51 78.54±17.34 00.00±00.00 G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 14 / 17
  • 15. Evaluation Discussion Preliminary evaluation showed that the method did not overcome the original version Large number of omission cases: a more conservative behavior Probably, due to the combination rules that returned a pooled BBA with a greater value in favor of ignorance the threshold employed to control the growth reduced the induction cases Results are not stable yet large standard deviation G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 15 / 17
  • 16. Conclusions Conclusion & Extensions An extension of TDT based on Dempster-Shafer Theory has been proposed and a preliminary evaluation has been made. The results are not very good, yet Extensions: Tackle the compexity of the model Pruning algorithms Consider other measures for the selection of the best candidates Further datasets should be considered Linked Data datasets G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 16 / 17
  • 17. Conclusions Thank you! Questions? G.Rizzo et al. (DIB- Univ. Aldo Moro) Evidence Terminological Decision Trees July 15 - 19, 2014 17 / 17