Evangelos Triantaphyllou

Louisiana State University | LSU · Division of Computer Science and Engineering

Background. Computer-aided diagnosis (CAD) can assist physicians in effective and efficient diagnostic decision-making. CAD systems are currently essential tools in some areas of clinical practice. In addition, it is one of the established fields of study in the interface of medicine and computer science. There are, however, still some critical cha...

Computer-aided diagnosis (CAD) in medicine is the result of a large amount of effort expended in the interface of medicine and computer science. As some CAD systems in medicine try to emulate the diagnostic decision-making process of medical experts, they can be considered as expert systems in medicine. Furthermore, CAD systems in medicine may proc...

Background Health utilities express the perceptions patients have on the impact potential adverse events of medical treatments may have on their quality of life. Being able to accurately assess health utilities is crucial when deciding what is the best treatment when multiple and diverse treatment options exist, or when performing a cost / utility...

Background: Shared decision making (SDM) for life-critical diseases or conditions is a crucial type of SDM. This type of SDM is still greatly underdeveloped and it faces a number of key challenges. The main goal of this study is to identify the challenges that impede the development and use of life-critical SDM. Methods: This is a hybrid research...

Pairwise comparison matrices (also known as Saaty matrices) in conjunction with a finite discrete scale for their quantification, are considered by many as efficient and effective means for eliciting personal preferences from decision makers. This study demonstrates that under a highly optimistic assumption, called the ultra-accurate decision maker...

People use knowledge on several cognitive tasks such as when they recognize objects, rank entities such as the alternatives in multi-criteria decision making, or for classification tasks of decision making / expert / intelligent systems. When people have sufficient relevant knowledge, they can make well-distinctive assessments among entities. Other...

Background Selecting the best treatment for life-critical conditions via a shared decision making approach is a uniquely important challenge. Besides data from the healthcare physicians, other data that need to be considered are the personal values and perceptions of the patient. Usually, these data come in the form of health-state utility values....

Traditional approaches to group decision making (GDM) problems for ranking a finite set of alternatives terminate when the experts involved in the GDM process reach a consensus. This paper proposes ways for analyzing the final results after a consensus has been reached in GDM. Results derived from this last step can be used to further enhance the u...

It is assumed that a group of experts is tasked to evaluate (rank) a finite set of alternatives during a group decision making (GDM) session. The GDM session may go through a number of iterations (stages) to reach a consensus. At each iteration at least one of the experts changes his/her ranking of some of the alternatives. The session terminates w...

The two publications (both in 2019): "A Systematic Survey of Computer-Aided Diagnosis in Medicine: Past and Present Developments," by Yanase and Triantaphyllou, 2019 and "The Seven Key Challenges for the Future of Computer-Aided Diagnosis in Medicine," by Yanase and Triantaphyllou, 2019, are a two-paper group closely related to CAD. They represen...

The two publications (both in 2019): "A Systematic Survey of Computer-Aided Diagnosis in Medicine: Past and Present Developments," by Yanase and Triantaphyllou, 2019 and "The Seven Key Challenges for the Future of Computer-Aided Diagnosis in Medicine," by Yanase and Triantaphyllou, 2019, are a two-paper group closely related to CAD. They represen...

Soon after our (e.g., by Kujawski E, Triantaphyllou E, and Yanase J) article entitled “Additive Multicriteria Decision Analysis Models: Misleading Aids for Life-Critical Shared Decision Making, May 2019, Medical Decision Making 39(4):437-499 was published, J Dolan published a “Letter to the Editor” in the same journal arguing against our main posit...

This is a Letter to the Editor accepted for publication in the J. of Medical Decision Making. It is written in response to another Letter to the Editor (by J. Dolan) which has argued against our main paper regarding the Misuse of MCDA models for Life-Critical Shared Decision Making (published in MDM).

Shared Decision Making (SDM) plays a fundamental role in patient-centered care. Life-Critical SDM is perhapsthe most critical area within SDM. However, there is a huge confusion of what kind of models may be used in life-critical SDM. Our recent paper (2019) published in the Journal of Medical Decision Making is the first significant effort for cre...

Background: There is growing interest in multicriteria decision analysis (MCDA) for shared decision making (SDM). A distinguishing feature is that a preferred treatment should extend years of life and/or improve health-related quality of life (HRQL). Additive MCDA models are inadequate for the task. A plethora of MCDA models exist, each claiming t...

This paper proposes a novel iterative approach for achieving consensus when a group of experts is given the task to rank a finite set of alternatives. Unlike traditional approaches which use various metrics to express expert disagreements, the proposed approach is based on a premetric concept to express such disagreements. This premetric approach c...

HOW TO IDENTIFY AND DEAL WITH NUMERICAL INCONSISTENCIES IN ELICITED HEALTH STATE UTILITIES ------------------------------------------------------------------------------- Yanase J 1, Triantaphyllou E 2 1Complete Decisions, LLC, Baton Rouge, LA, USA, 2Louisiana State University, Division of Computer Science & Engineering, Baton Rouge, LA, USA AB...

A system, method, and computer program product for evaluating a set of data records to identify critical records in said set which lie very close to a class boundary and are sensitive to small changes in attribute values, such that the small changes may result in the switching of classes. Additionally, a system, method, and computer program product...

In large databases, there may exist critical nuggets-small collections of records or instances that contain domain-specific important information. This information can be used for future decision making such as labeling of critical, unlabeled data records and improving classification results by reducing false positive and false negative errors. Thi...

Current classification algorithms usually do not try to achieve a balance between fitting and generalization when they infer models from training data. Furthermore, current algorithms ignore the fact that there may be different penalty costs for the false-positive, false-negative, and unclassifiable types. Thus, their performance may not be optimal...

A central problem in data mining is how to analyze observations grouped into two categories and infer some key patterns that may be implied by these observations. As discussed in Chapter 1, these observations describe different states of nature of the system or phenomenon of interest to the analyst.

Almost any use of a data mining and knowledge discovery method on a data set requires some discussion on the accuracy of the extracted model on some test data. This accuracy can be a general description of how well the extracted model classifies test data. Some studies split this accuracy rate into two rates: the false-positive and false-negative r...

The previous chapter studied the guided learning problem. In that setting, the analyst has the option to select which unclassified example to send to the oracle for classification and use that information to improve the understanding of the system under consideration. When the new example would unveil the need for an update, one had to use all t...

In many data mining and knowledge discovery applications a critical task is how to define the values of the various attributes that the analyst believes may be of significance. For easily quantifiable attributes (such as, age, weight, cost, etc.) this task is a rather straightforward one as it involves simple measurements and expressing the resul...

This chapter discusses a useful relationship between the CNF and DNF forms of the Boolean functions derivable from the same training data. This relationship can benefit approaches which attempt to solve large Boolean function inference problems and use either the CNF or the DNF form in representing a Boolean function.

The property of monotonicity has many applications. Its attractive mathematical advantages in inferring a model of the system of interest with high accuracy make the search for this property in data and its consecutive algorithmic exploitation, to be of high potential in data mining and knowledge discovery applications. The following developmen...

This chapter investigates the problem of classifying text documents into two disjoint classes. It does so by employing a data mining approach based on the OCAT algorithm. This chapter is based on the work discussed in [ Nieto Sanchez, Triantaphyllou, and Kraft, 2002]. In the present setting two sample sets of training examples (text document...

Most of the previous chapters discussed some application issues on a number of areas. This chapter discusses a case study in detail. The emphasis is on some comparative issues with other data mining techniques that do not use logic-based approaches. This chapter also provides a link to the data used in this study.

The previous two chapters discussed the development and key mathematical properties of some branch-and-bound (B&B) branch-and-bound (B&B) approaches for inferring a Boolean function in the form of a compact (i.e., with as few clauses as possible) CNF or DNF expression from two collections of disjoint examples. As was described in Chapters 2 and 3,...

This chapter is based on the findings presented in [Triantaphyllou and Soyster, 1996] and presents the motivation and definition of a special graph which can be easily derived from positive and negative examples. To understand the motivation for introducing this graph, consider a situation with n = 5 attributes.

This chapter discusses a revised branch-and-bound (B&B) algorithm for inferring a single clause (in CNF or DNF) from two disjoint sets of binary training examples. This algorithm is an extension of the B&B algorithm described in the previous chapter. Now the states of the search space are described by using more information and this seems to be cri...

In all previous discussions the problem was how to infer a general Boolean function based on some training examples. Such a Boolean function can be completely inferred if all possible binary examples (states) in the space of the attributes are used for training. Thus, one may never be 100% certain about the validity of the inferred knowledge when t...

For this case study we used a data set that described a number of clinical cases of breast cancer diagnoses. The data were divided into two disjoint sets of malignant and benign cases. We applied the OCAT approach, as it is embedded in the RA1 heuristic (see also Chapter 4), after the data were transformed into binary ones according to the method d...

In most of the previous treatments it was assumed that somehow we have available two disjoint sets of training data described by binary vectors, that is, the collections of the positive and negative examples. Then the problem was how to infer a Boolean function that “fits these data.” In other words, a Boolean function in CNF or DNF form that satis...

Mining of association rules from databases has attracted great interest because of its potentially very useful applications. Association rules are derived from a type of analysis that extracts information from coincidence [Blaxton and Westphal, 1998]. Sometimes called market basket analysis, this methodology allows a data analyst to discover correl...

Please see the four attached PDF files. This is a book.

Each of the previous chapters ends with a section with some concluding remarks tailored to the contents of the particular chapter. This section provides some comprehensive concluding remarks. As was mentioned earlier, there are many approaches to data mining and knowledge discovery from data sets. Such approaches include neural networks, closest ne...

Data mining and knowledge discovery is a family of computational methods that aim at collecting and analyzing data related to the function of a system of interest for the purpose of gaining a better understanding of it. This system of interest might be artificial or natural. According to the Merriam-Webster online dictionary the term system is deri...

Medical data mining has recently become one of the most popular topics in the data mining community. This is due to the societal importance of the field and also the particular computational challenges posed in this domain of data mining. However, current medical data mining approaches oftentimes use identical costs or just ignore them for the diff...

This paper presents the basics of a new paradigm that allows generators and consumers of global contextual information to determine an appropriate security level needed for contextual information. Security levels have a direct correlation with confidence in the integrity of contextual data and thus their processing. The new approach is based on the...

This communication focuses on a fundamental problem related to the recently introduced Reference-Dependent Regret Model (RDRM) [E. Kujawski, Syst Eng 8(2) (2005), 119– 137] for deterministic multi-criteria decision-making. Kujawski asserted that the RDRM model satisfies three properties. The first of these properties, referred to as the ‘‘independe...

The correct detection of welding flaws is important to the successful development of an automated weld inspection system. As a continuation of our previous efforts, this study investigates the performance of multi-layer perceptron (MLP) neural networks and case based reasoning (CBR) individually as well as their combined use. It is found that bette...

The Pima Indian diabetes (PID) dataset [1], originally donated by Vincent Sigillito from the Applied Physics Laboratory at the Johns Hopkins University, is one of the most well-known datasets for testing classification algorithms. This dataset consists of records describing 786 female patients of Pima Indian heritage which are at least 21 years old...

The main goal of the new field of data mining is the analysis of large and complex datasets. Some very important datasets may be derived from business and industrial activities. This kind of data is known as "enterprise data". The common characteristic of such datasets is that the analyst wishes to analyze them for the purpose of designing a more c...

The ELECTRE II and III methods enjoy a wide acceptance in solving multi-criteria decision-making (MCDM) problems. Research results in this paper reveal that there are some compelling reasons to doubt the correctness of the proposed rankings when the ELECTRE II and III methods are used. In a typical test we first used these methods to determine the...

Many classification studies often times conclude with a summary table which presents performance results of applying various data mining approaches on different datasets. No single method outperforms all methods all the time. Furthermore, the performance of a classiffication method in terms of its false-positive and false-negative rates may be tota...

Keywords Inference of Monotone Boolean Functions The Shannon Function and the Hansel Theorem Hansel Chains Devising a Smart Question-Asking Strategy Conclusions See also References

Keywords Overview of Automatic Classification of Documents Examples of Optimization in Document Classification Optimization in the Principle of Least Effort Optimization in the Vector Space Model Optimization in the Classification of Text Documents Conclusions and Future Research See also References

Keywords Background Information Optimization Approaches Concluding Remarks See also References

Unlocking the mystery of natural phenomena is a universal objective in scientific research. The rules governing a phenomenon can most often be learned by observing it under a sufficiently large number of conditions that are sufficiently high in resolution. The general knowledge discovery process is not always easy or efficient, and even if knowledg...

This chapter considers four key data mining areas which seem to have a promising future. These areas are: web mining, visual data mining, text data mining, and distributed data mining. The reason of their importance is to be found in the valuable applications they can support but also in the proliferation of the web and in the dramatic improvements...

This chapter reviews a data mining and knowledge discovery approach called OCAT (for One Clause At a Time). The OCAT approach is based on concepts of mathematical logic and discrete optimization. As input it uses samples of the performance of the system (or phenomenon) under consideration and then it extracts its underlying behavior in terms of a c...

This book will give the reader a perspective into the core theory and practice of data mining and knowledge discovery (DM&KD). Its chapters combine many theoretical foundations for various DM&KD methods, and they present a rich array of examples—many of which are drawn from real-life applications. Most of the theoretical developments discussed are...

This chapter first introduces the main issues of multi-criteria decision analysis (MCDA). This involves discussion on some well-known MCDA methods. Next it describes some ranking irregularities when some MCDA methods are used. Ranking irregularities occur when certain manipulations on the structure of a simple MCDA problem are performed. Though a p...

Multicriteria decision analysis (MCDA) problems (also known as multicriteria decision-making or MCDM) involve the ranking of a finite set of alternatives in terms of a finite number of decision criteria. Often times such criteria may be in conflict with each other. That is, an MCDA problem may involve both benefit and cost criteria at the same time...

The purpose of this research is to develop a depot location model to manage the resources needed for efficient and economic power restoration in an area which has experienced an outage. Determining the optimal number of depots, optimal location of depots, and the optimal number of repair crews and/or equipment is of significant importance in power...

This paper addresses the problem of minimizing the average query complexity of inferring a pair of nested monotone Boolean functions defined on {0,1}n using a pair of oracles. Here, nested refers to the case when one of the functions is always greater than or equal to the other function. It is shown that the nested case is equivalent to inferring t...

Mining association rules from databases has attracted great interest because of its potentially very practical applications. Given a database, the problem of interest is how to mine association rules (which could describe patterns of consumers' behaviors) in an efficient and effective way. The databases involved in today's business environments can...

This paper introduces an incremental algorithm for learning a Boolean function from examples. The functions are constructed in the disjunctive normal form (DNF) or the conjunctive normal form (CNF) and emphasis is placed in inferring functions with as few clauses as possible. This incremental algorithm can be combined with any existing algorithm th...

This paper proposes a new approach for classifying text documents into two disjoint classes. The new approach is based on extracting patterns, in the form of two logical expressions, which are defined on various features (indexing terms) of the documents. The pattern extraction is aimed at providing descriptions (in the form of two logical expressi...

his paper addresses the problem of completely reconstructing deterministic monotone Boolean functions via membership queries. The minimum average query complexity is guaranteed via recursion, where partially ordered sets (posets) make up the overlapping subproblems. For problems with up to 4 variables, the posets' optimality conditions are summariz...

One of the most crucial steps in many multicriteria decision making methods (MCDM) is the accurate estimation of the pertinent data [18]. Very often these data cannot be known in terms of absolute values. For instance, what is the worth of the ith alternative in terms of a political impact criterion? Although information about questions like the pr...

This paper compares the total inventory costs (TIC) of five lot-sizing techniques. The add-drop heuristic (ADH) is a capacitated technique and the lot-for-lot (L4L), fixed period quantity (FPQ), least unit cost (LUC) and the silver-meal heuristic (SMH) are uncapacitated techniques. The TIC is considered as a function of the reorder interval (RI). T...

Many researchers have long observed some cases in which certain ranking irregularities can occur when the original analytic hierarchy process (AHP), or some of its variants, are used. This paper presents two new categories of ranking irregularities which defy common intuition. These ranking irregularities occur when one decomposes a decision proble...

The goal in a classification problem is to uncover a system that places examples into two or more mutually exclusive groups. Identifying a classification system is beneficial in several ways. First of all, examples can be organized in a meaningful way, which will make the exploration and retrieval of examples belonging to specific group(s) more eff...

In many decision problems the focus is on ranking a set of m alternatives in terms of a number, say n, of decision criteria. Given are the performance values of the alternatives for each one of the criteria and the weights of importance of the criteria. This paper demonstrates that if one assumes that the criteria weights are changeable, then the n...

Using fuzzy c-means as the data-mining tool, this study evaluates the effectiveness of sampling methods in producing the knowledge of interest. The effectiveness is shown in terms of the representative-ness of sampling data and both the accuracy and errors of sampled data sets when subjected to the fuzzy clustering algorithm. Two population data in...

This paper introduces a number of reliability criteria for computer-aided diagnostic systems for breast cancer. These criteria are then used to analyze some published neural network systems. It is also shown that the property of monotonicity for the data is rather natural in this medical domain, and it has the potential to significantly improve the...

The analysis of the way people make decisions (prescriptive theories) or the way people ought to make decisions (normative theories) is perhaps as old as the recorded history of mankind. Of course, not all these analyses were characterized by the rigorous scientific approaches we see in the literature today. Therefore, it is not surprising that the...

Multi-Criteria Decision Making (MCDM) has been one of the fastest growing problem areas in many disciplines. The central problem is how to evaluate a set of alternatives in terms of a number of criteria. Although this problem is very relevant in practice, there are few methods available and their quality is hard to determine. Thus, the question `Wh...

It is widely accepted today that people do not always behave the way the well studied normative theories say they ought to behave (see, for instance, [Allais and Hagen, 1979], [Bell, et al., 1988], [Ellsberg, 1961], and Raiffa [1984]. Many decision theories (especially game theories) assume that the decision makers are always perfectly rational. To...

As it was discussed in previous chapters, an appealing approach for eliciting qualitative data for an MCDM problem is to use pairwise comparisons. Next suppose that a decision maker wishes to elicit the relative priorities, or weights of importance, of n entities via a sequence of pairwise comparisons. As before, these n entities could be the decis...

For a long time it has been recognized that an exact description of many real life physical situations may be virtually impossible. This is due to the high degree of imprecision involved in real world situations. Zadeh, in his seminal papers [Zadeh, 1965; and 1968], proposed fuzzy set theory as the means for quantifying the inherent fuzziness that...

The MCDM methods presented in the second chapter are among the most widely used ones. As it will be seen in the following sections, however, the part of the methods that processes a decision matrix (i.e., step 3 in Section 2.2) may give different answers to the same problem. Because only the WPM, the AHP, the revised AHP, and the TOPSIS method are...

As it was described in the previous chapters, pairwise comparisons play an important role in MCDM problems. They often provide an effective and efficient manner for eliciting qualitative information from the decision maker(s). However, a severe drawback of their application is the often large number of them. If there are n objects (also called enti...

In this chapter four deterministic MCDM methods of the ones presented in the second chapter are fuzzified. These are the WSM, the WPM, the AHP (original and ideal mode), and the TOPSIS method. The ELECTRE is not examined (since the TOPSIS method seems to be superior to it). The multiplicative AHP (as described in Section 11.4) is not studied either...

As it was mentioned in Chapter 1, the typical problem examined by the AHP consists of a set of alternatives and a set of decision criteria. Since this problem is very common in many engineering, science, and economic applications, the AHP has been a very popular decision tool. Another reason which contributed to the wide use of the AHP in such appl...

Chapter 9 presented an evaluation of the AHP, the revised AHP, the WPM, and the TOPSIS methods in terms of two evaluative criteria. Another evaluation was presented in Chapter 10 for the AHP and the revised AHP methods. The issue of evaluating MCDM methods is a controversial one in the decision analysis / decision making communities. This chapter i...

With the continuing proliferation of decision methods and their variants, it is important to have an understanding of their comparative value. Each of the methods uses numeric techniques to help decision makers choose among a discrete set of alternative decisions. This is achieved on the basis of the impact of the alternatives on certain criteria a...

The first step in any MCDM problem is to define the set of alternatives and the set of decision criteria that the alternatives need to be evaluated with. Although this is an enormously critical step, its formulation cannot easily be captured with a standard modeling procedure. This task appeals more to the art aspect of MCDM than to the science one...

As it was mentioned in Chapter 3, an important issue in MCDM methods is to be able to determine the relative weights of importance of a collection of entities (such as the alternatives to be studied in terms of a single decision criterion). This task is similar and closely related to the problem of determining the degree of membership of the elemen...