Skip to main content

Über dieses Buch

This book constitutes the proceedings of the 15th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2018, held in Mallorca, Spain, in October 2018.

The 24 papers presented in this volume were carefully reviewed and selected from 43 submissions. The book also contains one invited talk in full paper length. The papers were organized in topical sections named: aggregation operators, fuzzy measures and integrals; decision making; clustering and classification; and data privacy and security.



Correction to: Modeling Decisions for Artificial Intelligence

The original versions of chapters “Graded Logic Aggregation” and “Implicative Weights as Importance Quantifiers in Evaluation Criteria” have been revised; minor errors in the text have been corrected at the request of the author.
Vicenç Torra, Yasuo Narukawa, Isabel Aguiló, Manuel González-Hidalgo

Invited Paper


Graded Logic Aggregation

This paper summarizes basic properties of graded logic – a natural soft computing generalization of classical Boolean logic. Using graded logic aggregators we can build evaluation criteria and apply them in decision engineering. This paper is an extended summary that surveys key concepts of graded logic and graded logic aggregation.
Jozo Dujmović

Aggregation Operators, Fuzzy Measures and Integrals


Coherent Risk Measures Derived from Utility Functions

Coherent risk measures in financial management are discussed from the view point of average value-at-risks with risk spectra. A minimization problem of the distance between risk estimations through decision maker’s utility and coherent risk measures with risk spectra is introduced. The risk spectrum of the optimal coherent risk measures in this problem is obtained and it inherits the risk averse property of utility functions. Various properties of coherent risk measures and risk spectrum are demonstrated. Several numerical examples are given to illustrate the results.
Yuji Yoshida

On k--additive Aggregation Functions

To generalize the concept of k-maxitivity and k-additivity, we introduce k-\(\oplus \)-additive aggregation functions. We also characterize this kind of aggregation functions under some special conditions. Several examples are given to illustrate the new definitions.
Fateme Kouchakinejad, Anna Kolesárová, Radko Mesiar

Constructing an Outranking Relation with Weighted OWA for Multi-criteria Decision Analysis

Some decision aiding methods are based on constructing and exploiting outranking relations. An alternative a outranks another b if a is at least as good as b (aSb). One well known method in this field is ELECTRE. The outranking relation is usually built by means of a weighted average (WA) of the votes given by a set of criterion with respect to the fulfilment of aSb. The value obtained represent the strength of the majority opinion. The WA operator can be observed to have sometimes an undesired compensative effect. In this paper we propose the use of other aggregation operators with different mathematical properties. In particular, we substitute the WA by three operators from the Ordered Weighted Average (OWA) family of operators because it permits to decide the degree of andness/orness that is used during the aggregation. The OWAWA (Ordered Weighted Average Weighted Average), WOWA (Weighted Ordered Weighted Average) and IOWA (Induced Ordered Weighted Average) operators are studied. They are capable to combine the importance given to each criterion with the conjunctive/disjunctive requirement applied in the definition of the outranking relation.
Jonathan Ayebakuro Orama, Aida Valls

Sugeno Integrals and the Commutation Problem

In decision problems involving two dimensions (like several agents and several criteria) the properties of expected utility ensure that the result of a multicriteria multiperson evaluation does not depend on the order with which the aggregations of local evaluations are performed (agents first, criteria next, or the converse). We say that the aggregations on each dimension commute. Ben Amor, Essghaier and Fargier have shown that this property holds when using pessimistic possibilistic integrals on each dimension, or optimistic ones, while it fails when using a pessimistic possibilistic integral on one dimension and an optimistic one on the other. This paper studies and completely solves this problem when Sugeno integrals are used in place of possibilistic integrals, indicating that there are capacities other than possibility and necessity measures that ensure commutation of Sugeno integrals.
Didier Dubois, Hélène Fargier, Agnès Rico

Characterization of k-Choquet Integrals

In the present paper we characterize the class of all n-ary k-Choquet integrals and we find a minimal subset of points in the unit hypercube, the values on which fully determine the k-Choquet integral.
L’ubomíra Horanská, Zdenko Takáč

Event-Based Transformations of Set Functions and the Consensus Requirement

Non-additive measures, capacities or generally set functions are widely used in decision models, data processing and game theory. In these applications we can find many structures identified as linear transformations or linear operators. The most remarkable of them are Choquet integral, Möbius transform, interaction transform, Shapley value. The main goal of the presented paper is to study some of them recently called event-based linear transformations. We describe them considering the set of all possible linear operators as a linear space w.r.t. their linear combinations and compute the dimensions of its some subspaces. We also study the consensus requirement, i.e. we analyze the condition when the linear operator maps one family of non-additive measures to other family.
Andrey G. Bronevich, Igor N. Rozenberg

Association Analysis on Interval-Valued Fuzzy Sets

The aim of this paper is to generalize the concept of association rules for interval-valued fuzzy sets. Interval-valued fuzzy sets allow for intervals of membership degrees to be assigned to each element of the universe. These intervals may be interpreted as partial information where the exact membership degree is not known. The paper provides a generalized definition of support and confidence, which are the most commonly known measures of quality of a rule.
Petra Murinová, Viktor Pavliska, Michal Burda

Fuzzy Hit-or-Miss Transform Using Uninorms

The Hit-or-Miss transform (HMT) is a morphological operator which has been successfully used to identify shapes and patterns satisfying certain geometric restrictions in an image. Recently, a novel HMT operator, called the fuzzy morphological HMT, was introduced within the framework of the fuzzy mathematical morphology based on fuzzy conjunctions and fuzzy implication functions. Taking into account that the particular case of considering a t-norm as fuzzy conjunction and its residual implication as fuzzy implication functions has proved its potential in several applications, in this paper, the case when residual implications derived from uninorms and a general fuzzy conjunction, possibly a t-norm or the same uninorm, is deeply analysed. In particular, some theoretical results related to properties desirable for the applications are proved. Finally, some experimental results are presented showing the potential of this choice of operator to detect shapes and patterns in images.
Pedro Bibiloni, Manuel González-Hidalgo, Sebastia Massanet, Arnau Mir, Daniel Ruiz-Aguilera

Learning Fuzzy Measures for Aggregation in Fuzzy Rule-Based Models

Fuzzy measures are used to express background knowledge of the information sources. In fuzzy rule-based models, the rule confidence gives an important information about the final classes and their relevance. This work proposes to use fuzzy measures and integrals to combine rules confidences when making a decision. A Sugeno \(\lambda \)-measure and a distorted probability have been used in this process. A clinical decision support system (CDSS) has been built by applying this approach to a medical dataset. Then we use our system to estimate the risk of developing diabetic retinopathy. We show performance results comparing our system with others in the literature.
Emran Saleh, Aida Valls, Antonio Moreno, Pedro Romero-Aroca, Vicenç Torra, Humberto Bustince

Decision Making


Extraction of Patterns to Support Dairy Culling Management

The management of a dairy farm involves taking decisions such as culling a subset of cows to improve the dairy production. Culling is the departure of cows from the herd due to sale, slaughter or death. Commonly the culling process is based on the farmer experience but there is not a general procedure to carry it out. In the present paper we use both, a method based on indistinguishability relations and the anti-unification concept, to extract patterns that characterise the cows according to their average milk production of the first lactation. Our goal is to identify as soon as possible poorly productive cows during her first lactation, which may be candidates to be culled.
M. López-Suárez, E. Armengol, S. Calsamiglia, L. Castillejos

An Axiomatisation of the Banzhaf Value and Interaction Index for Multichoice Games

We provide an axiomatisation of the Banzhaf value (or power index) and the Banzhaf interaction index for multichoice games, which are a generalisation of cooperative games with several levels of participation. Multichoice games can model any aggregation model in multicriteria decision making, provided the attributes take a finite number of values. Our axiomatisation uses standard axioms of the Banzhaf value for classical games (linearity, null axiom, symmetry), an invariance axiom specific to the multichoice context, and a generalisation of the 2-efficiency axiom, characteristic of the Banzhaf value.
Mustapha Ridaoui, Michel Grabisch, Christophe Labreuche

Fuzzy Positive Primitive Formulas

Can non-classical logic contribute to the analysis of complexity in computer science? In this paper, we give a step towards the solution of this open problem, taking a logical model-theoretic approach to the analysis of complexity in fuzzy constraint satisfaction. We study fuzzy positive-primitive sentences, and we present an algebraic characterization of classes axiomatized by this kind of sentences in terms of homomorphisms and finite direct products. The ultimate goal is to study the expressiveness and reasoning mechanisms of non-classical languages, with respect to constraint satisfaction problems and, in general, in modelling decision scenarios.
Pilar Dellunde

Basic Level Concepts as a Means to Better Interpretability of Boolean Matrix Factors and Their Application to Clustering

We present an initial study linking in cognitive psychology well known phenomenon of basic level concepts and a general Boolean matrix factorization method. The result of this fusion is a new algorithm producing factors that explain a large portion of the input data and that are easy to interpret. Moreover, the link with the cognitive psychology allowed us to design a new clustering algorithm that groups objects into clusters that are close to human perception. In addition we present experiments that provide insight to the relationship between basic level concepts and Boolean factors.
Petr Krajča, Martin Trnecka

Fuzzy Type Powerset Operators and F-Transforms

We introduce two types of aggregation operators for lattice-valued fuzzy sets, called fuzzy type powerset operators and fuzzy type F-transforms, which are derived from classical powerset operators and F-transforms, respectively. We prove that, in contrast with classical powerset operators, fuzzy type powerset operators form a subclass of fuzzy type F-transforms. Some examples of fuzzy type powerset operators are presented.
Jiří Močkoř

Implicative Weights as Importance Quantifiers in Evaluation Criteria

This paper investigates properties of implicative weights and the use of implicative weights in evaluation criteria. We analyze and compare twelve different forms of implication and compare them with multiplicative weights and exponential weights that are also used in evaluation criteria. Since weighted conjunction is based on implicative weights, we also investigate the usability of weighted conjunction in evaluation criteria.
Vicenç Torra

Balancing Assembly Lines and Matching Demand Through Worker Reallocations

Assembly lines are of great importance in most actual production systems and thus continue attracting strong research interest. We address a real industry scenario where the aim of the line is to target a production output that meets, as much as possible, a given demand forecast. To the best of our knowledge, the existing literature has not tackled this problem, and we named it the demand-driven assembly line (re)balancing problem. A mixed integer programming model is developed, solved using genetic algorithm, and tested in the straight assembly line, providing useful insights about the dynamics of worker reallocations.
Randall Mauricio Pérez-Wheelock, Van-Nam Huynh

Clustering and Classification


Optimal Clustering with Twofold Memberships

This paper proposes two clustering algorithms of twofold memberships for each cluster. One uses a membership similar to that in K-means, while another membership is defined for a core of a cluster, which is compared to the lower approximation of a cluster in rough K-means. Two ideas for the lower approximation are proposed in this paper: one uses a neighborhood of a cluster boundary and another uses a simple circle from a cluster center. By using the two memberships, two alternate optimization algorithms are proposed. Numerical examples show the effectiveness of the proposed algorithms.
Sadaaki Miyamoto, Jong Moon Choi, Yasunori Endo, Van Nam Huynh

Privacy Preserving Collaborative Fuzzy Co-clustering of Three-Mode Cooccurrence Data

Co-cluster structure analysis with three-mode cooccurrence information is a potential approach in summarizing multi-source relational data in such tasks as user-product purchase history analysis. This paper proposes a privacy preserving framework for jointly performing three-mode fuzzy co-clustering under collaboration among two organizations, which independently store object-item cooccurrence information and item-ingredient cooccurrence information, respectively. Even when they cannot mutually share elements of the cooccurrence matrices, the intrinsic co-cluster structures are revealed without publishing each elements of relational data but sharing only the structural information.
Katsuhiro Honda, Shotaro Matsuzaki, Seiki Ubukata, Akira Notsu

Generalized Fuzzy c-Means Clustering and Its Theoretical Properties

This study shows that a generalized fuzzy c-means (gFCM) clustering algorithm, which covers standard fuzzy c-means clustering, can be constructed if a given fuzzified function, its derivative, and its inverse derivative can be calculated. Furthermore, our results show that the fuzzy classification function for gFCM exhibits similar behavior to that of standard fuzzy c-means clustering.
Yuchi Kanzawa, Sadaaki Miyamoto

A Self-tuning Possibilistic c-Means Clustering Algorithm

Most c-means clustering models have serious difficulties when facing clusters of different sizes and severely outlier data. The possibilistic c-means (PCM) algorithm can handle both problems to some extent. However, its recommended initialization using a terminal partition produced by the probabilistic fuzzy c-means does not work when severe outliers are present. This paper proposes a possibilistic c-means clustering model that uses only three parameters independently of the number of clusters, which is able to more robustly handle the above mentioned obstacles. Numerical evaluation involving synthetic and standard test data sets prove the advantages of the proposed clustering model.
László Szilágyi, Szidónia Lefkovits, Zsolt Levente Kucsván

k-CCM: A Center-Based Algorithm for Clustering Categorical Data with Missing Values

This paper focuses on solving the problem of clustering for categorical data with missing values. Specifically, we design a new framework that can impute missing values and assign objects into appropriate clusters. For the imputation step, we use a decision tree-based method to fill in missing values. For the clustering step, we use a kernel density estimation approach to define cluster centers and an information theoretic-based dissimilarity measure to quantify the differences between objects. Then, we propose a center-based algorithm for clustering categorical data with missing values, namely k-CCM. An experimental evaluation was performed on real-life datasets with missing values to compare the performance of the proposed algorithm with other popular clustering algorithms in terms of clustering quality. Generally, the experimental result shows that the proposed algorithm has a comparative performance when compared to other algorithms for all datasets.
Duy-Tai Dinh, Van-Nam Huynh

Data Privacy and Security


WEDL-NIDS: Improving Network Intrusion Detection Using Word Embedding-Based Deep Learning Method

A Network Intrusion Detection System (NIDS) helps system administrators to detect security breaches in their organization. Current research focus on machine learning based network intrusion detection method. However, as numerous complicated attack types have growingly appeared and evolved in recent years, obtaining high detection rates is increasingly difficult. Also, the performance of a NIDS is highly dependent on feature design, while a feature set that can accurately characterize network traffic is still manually designed and usually costs lots of time. In this paper, we propose an improved NIDS using word embedding-based deep learning (WEDL-NIDS), which has the ability of dimension reduction and learning features from data with sophisticated structure. The experimental results show that the proposed method outperforms previous methods in terms of accuracy and false alarm rate, which successfully demonstrates its effectiveness in both dimension reduction and practical detection ability.
Jianjing Cui, Jun Long, Erxue Min, Yugang Mao

Anonymization of Unstructured Data via Named-Entity Recognition

The anonymization of structured data has been widely studied in recent years. However, anonymizing unstructured data (typically text documents) remains a highly manual task and needs more attention from researchers. The main difficulty when dealing with unstructured data is that no database schema is available that can be used to measure privacy risks. In fact, confidential data and quasi-identifier values may be spread throughout the documents to be anonymized. In this work we propose to use a named-entity recognition tagger based on machine learning. The ultimate aim is to build a system capable of detecting all attributes that have privacy implications (identifiers, quasi-identifiers and sensitive attributes). In particular, we present a proof of concept focused on the detection of confidential attributes. We consider a case study in which confidential values to be detected are disease names in medical diagnoses. Once these confidential attribute values are located, one can use standard statistical disclosure control techniques for structured data to control disclosure risk.
Fadi Hassan, Josep Domingo-Ferrer, Jordi Soria-Comas

On the Application of SDC Stream Methods to Card Payments Analytics

Banks and financial services have to constantly innovate their online payment services to avoid large digital companies take the control of online card transactions, relegating traditional banks to simple payments carriers. Apart from creating new payment methods (e.g. contact-less cards, mobile wallets, etc.), banks offers new services based on historical payments data to endow traditional payments methods with new services and functionalities. In this latter case, it is where privacy preserving techniques play a fundamental role ensuring personal data is managed full-filling all the applicable laws and regulations. In this paper, we introduce some ideas about how SDC stream anonymization methods could be used to mask payments data streams. Besides, we also provide some experimental results over a real card payments database.
Miguel Nuñez-del-Prado, Jordi Nin


Weitere Informationen

Premium Partner