Ph.D., December of 2008. Committee Chairman.
(click here to access Xiaoting's personal webpage at LSU)
Ph.D. Dissertation Title:
A Study of Regret and Rejoicing and a New MCDM Method Based on Them
ABSTRACT:
Multi-criteria decision-making (MCDM) is one of the most widely used decision methodologies in the sciences, business, and engineering worlds. MCDM methods aim at improving the quality of decisions by making the process more explicit, rational, and efficient. One controversial problem is that some well-known MCDM methods, like the additive AHP methods and the ELECTRE II and III methods, may cause some types of rank reversal problems. Rank reversal means that the ranking between two alternatives might be reversed after some variation occurs to the decision problem, like adding a new alternative, dropping an old alternative or replacing a non-optimal alternative by a worse one etc. Usually such a rank reversal is undesirable for decision-making problems. If a method does allow it to happen, the validity of the method could be questioned. However, some recent studies indicate that rank reversals could also happen because of people’s rational preference reversal which may be caused by their emotional feelings, like regret and rejoicing.
Since regret and rejoicing may play a pivotal role in evaluating alternatives in MCDM problems, sometimes the decision maker (DM) may want to anticipate these emotional feelings and consider them in the decision-making process. Most of the regret models in the literature use continuous functions to measure this emotional factor. This dissertation proposes to use an approach based on a linguistic scale and pairwise comparisons to measure a DM’s anticipated regret and rejoicing feelings. The approach is shown to exhibit some key advantages over existing approaches. Next a multiplicative MCDM model is adopted to aggregate the alternatives’ associated regret and rejoicing values with their performance values to get their final priorities and then rank them. A simulated numerical example is used to illustrate the process of the proposed method. Some sensitivity analyses which aim at examining how changes of regret and rejoicing values might affect the ranking results of the decision problems are also developed. Then a fuzzy version of the new method is introduced and illustrated by a numerical example. Finally, some concluding remarks are made. Ranking intransitivity and some other issues about the proposed method are analyzed too.
KEY WORDS: rank multi-criteria decsion making, reversals, regret theory, rejoicing, pairwise comparisons, sensitivity analysis
Click here to access this Ph.D. at LSU's Thesis/Dissertation electronic repository
Current Employment:
E-mail: XWang8@lsu.edu
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
3 number of refereed journal paper
2 chapters in edited books
A number of conference presentations
M.S. in Systems Science, May of 2007. Committee Chairman.
M.S. Thesis Title:
Data Exploration by Using the Monotonicity Property
ABSTRACT:
Dealing with different misclassification costs has been a big problem for classification. Some algorithms can predict quite accurately when assuming the misclassification costs for each class are the same, like most rule induction methods. However, when the misclassification costs change, which is a common phenomenon in reality, these algorithms are not capable of adjusting their results. Some other algorithms, like the Bayesian methods, have the ability to yield probabilities of a certain unclassified example belonging to given classes, which is helpful to make modification on the results according to different misclassification costs. The shortcoming of such algorithms is, when the misclassification costs for each class are the same, they do not generate the most accurate results.
This thesis attempts to incorporate the merits of both kinds of algorithms into one. That is, to develop a new algorithm which can predict relatively accurately and can adjust to the change of misclassification costs.
The strategy of the new algorithm is to create a weighted voting system. A weighted voting system will evaluate the evidence of the new example belonging to each class, calculate the assessment of probabilities for the example, and assign the example to a certain class according to the probabilities as well as the misclassification costs.
The main problem of creating a weighted voting system is to decide the optimal weights of the individual votes. To solve this problem, we will mainly refer to the monotonicity property. People have found the monotonicity property does not only exist in pure monotone systems, but also exists in non-monotone systems. Since the study of the monotonicity property has been a huge success on monotone systems, it is only natural to apply the monotonicity property to non-monotone systems too.
This thesis deals only with binary systems. Though such systems hardly exist in practice, this treatment provides concrete ideas for the development of general solution algorithms.
After the final algorithm has been formulated, it has been tested on a wide range of randomly generated synthetic datasets. It has also been compared with other existing classifiers. The results indicate this algorithm performs both effectively and efficiently.
KEY WORDS: Boolean function, binary systems, weighted voting system, monotonicity property, probabilities, misclassification cost, classifier, data mining.
Click here to access this MS Thesis at LSU's Thesis/Dissertation electronic repository
M.S., Summer of 2007. Committee Chairman.
M.S. Thesis Title:
Study of Ranking Irregularities When Evaluating Alternatives by Using Some ELECTRE Methods and a Proposed New MCDM Method Based on Regret and Rejoicing
ABSTRACT:
Multi-criteria decision-making (MCDM) is one of the most widely used decision methodologies in the sciences, business, and engineering worlds. MCDM methods aim at improving the quality of decisions by making the decision-making process more explicit, rational, and efficient. Some current applications of MCDM include the use on portfolio management, evaluation of technology investment decisions, flood management for flood hazard mitigation and allocation of scarce homeland security resources in pursuit of economic efficiency etc.
Though MCDM has attracted the interest of researchers and practitioners for many years in a wide spectrum of areas, it is far from being mature and there are still a lot of unsolved issues. One intriguing problem with MCDM methods is that oftentimes different methods may yield different answers when they are fed with exactly the same decision problem. Thus, the issue of evaluating the relative performance of different MCDM methods is naturally raised. Since it is practically impossible to know which alternative is the best for a given decision problem, some kind of testing procedures need to be developed. One such procedure is to examine the stability of an MCDM method’s mathematical process by checking the validity of its proposed rankings. The origin of this procedure comes from some studies where the original additive AHP method, one of the well-known MCDM methods, was found to allow some rank reversal problems to happen.
The ELECTRE method is another type of well-known MCDM method. Among different variants of the ELECTRE method, the ELECTRE II and III methods have been widely accepted in solving MCDM problems in the engineering world. Applications include the assessment of complex civil engineering projects, site selection for the disposal of nuclear waste and building a new nuclear plant etc. Though there have been so many applications of these two methods, the ELECTRE II and III methods have never been studied in detail for the validity of their proposed rankings. Thus, the aim of this thesis is first to examine if these two methods suffer of any type of ranking irregularity problems and analyze the reasons of the phenomenon.
As the research results in this thesis revealed, the ELECTRE II and III methods do allow some types of ranking irregularities to happen. In a typical test these two methods were first used to determine the best alternative for a given MCDM problem. Next a non-optimal alternative was randomly replaced by a worse one and the new alternative set was ranked again without changing any of the other data. The computational tests revealed that sometimes the ELECTRE II and III methods might change the indication of the best alternative. That is, rank reversals may happen with these two methods. The two methods are also evaluated in terms of two other ranking tests and they failed them as well. Two real-life cases are described to demonstrate the occurrence of rank reversals. Then reasons behind the phenomenon are analyzed in detail. Next an empirical study and some real-life case studies were executed and discussed. The results of these examinations show that the rates of those ranking irregularities were rather significant in both the simulated decision problems and the real-life cases studied in this research.
Usually the problem of rank reversal is undesirable for decision-making problems. If a method does allow it to happen because of its own mathematical instabilities/defects, the validity of the method could be questioned, like the ELECTRE II and III methods and the AHP method. However, some recent studies showed that rank reversals could also happen because people may reverse their preference due to some strong emotional feelings, like regret and rejoicing. The notion of regret and rejoicing comes from the fact that humans often base their choices on comparisons across the alternatives under consideration and relative to “what might have been” under another choice. Since regret and rejoicing may play a pivotal role in evaluating alternatives in MCDM problems, one needs to properly assess their impact in the decision making process. In the Chapter 8, how the feelings of regret and rejoicing may cause people to reverse their preference is described. Furthermore, some ongoing studies on a new regret and rejoicing based MCDM method with some essential properties are discussed.
KEY WORDS: Multi-Criteria Decision Making, ranking reversals, ELECTRE methods.
Current Employment:
E-mail: XWang8@lsu.edu
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
1 refereed journal paper
2 chapters in edited books
A number of conference presentations
M.S., May of 2002. Committee Chairman.
M.S. Thesis Title:
The Effect of Initial Selections in Estimating the Missing Comparisons in an Incomplete AHP Matrix
ABSTRACT:
The Analytic Hierarchy Process (AHP) is both easy to understand and very versatile with its technique of comparing the alternatives by means of sequence of pairwise comparison matrices. This paper addresses two problems concerning the incomplete AHP. The first problem is to determine which initial comparisons should be asked as the starting point. Using the minimal number of initial comparisons (i.e., comparisons), five different selection strategies will be investigated. The second problem of this paper is to determine the optimal number of the initial comparisons. The results show that these selection strategies will influence the accuracy in estimating the missing comparisons. Furthermore, the minimal number of initial comparisons may be sufficient to determine the weight vector with relative precision.
KEY WORDS: Analytical Hierarchy Process, Incomplete AHP Matrix, Initial Comparisons.
Current Employment:
E-mail: ssetiaw@lsu.edu
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
1 pending refereed journal paper
Graduated with the Ph.D. in Engineering Science with focus area on Industrial Engineering (May 2002), Ph.D. Committee Chairman.
Ph.D. Dissertation Title:
Data Mining and Knowledge Discovery: A Guided Approach Based on Monotone Boolean Functions
ABSTRACT:
This dissertation deals with an important problem in Knowledge Discovery and Data Mining (KD & DM), and Information Technology (IT) in general. It addresses the problem of efficiently learning monotone Boolean functions via membership queries to oracles. The monotone Boolean function can be thought of as a phenomenon, such as breast cancer or a computer crash, together with a set of predictor variables. The oracle can be thought of as an entity that knows the underlying monotone Boolean function, and provides a Boolean response to each query. In practice, it may take the shape of a human expert, or it may be the outcome of performing tasks such as running experiments or searching large databases.
Monotone Boolean functions have a general knowledge representation power and are inherently frequent in applications. A key goal of this dissertation is to demonstrate the wide spectrum of important real-life applications that can be analyzed by using the new proposed computational approaches. The applications of breast cancer diagnosis, computer crashing, college acceptance policies, and record linkage in databases are here used to demonstrate this point and illustrate the algorithmic details. Monotone Boolean functions have the added benefit of being intuitive. This property is perhaps the most important in learning environments, especially when human interaction is involved, since people tend to make better use of knowledge they can easily interpret, understand, validate, and remember.
The main goal of this dissertation is to design new algorithms that can minimize the average number of queries used to completely reconstruct monotone Boolean functions defined on a finite set of vectors V = {0,1}n. The optimal query selections are found via a recursive algorithm in exponential time (in the size of V). The optimality conditions are then summarized in the simple form of evaluative criteria, which are near optimal and only take polynomial time to compute. Extensive unbiased empirical results show that the evaluative criterion approach is far superior to any of the existing methods. In fact, the reduction in average number of queries increases exponentially with the number of variables n, and faster than exponentially with the oracle's error rate.
CONGRATULATIONS to Dr. Vetle I. Torvik (Ph.D. in Engineering Science at LSU in May 2002) for winning the LSU Alumni Distinguished Dissertation Award for 2002 (the highest Ph.D. Dissertation Award at LSU). Please note that this the second Ph.D. Dissertation award for Vetle this year!
CONGRATULATIONS to Dr. Vetle I. Torvik (Ph.D. in Engineering Science at LSU in May 2002) for winning the College of Engineering Exemplary Dissertation Award for 2002 (click here to see this award). A job well done Vetle!
Current Employment:
Dr. Torvik is currently (Fall of 2001) in the Psychiatry Institute, College of Medicine, University of Illinois at Chicago. His address is:
Vetle I. Torvik, Ph.D
Psychiatric Institute MC 912
University of Illinois at Chicago
1601 W Taylor St
Chicago, IL 60612
U.S.A.
He works there as a Research Assistant Professor in Neuroinformatics on a 5-year project funded by NIH. This project is related to the application of data mining and knowledge discovery approaches to some medical problems. More details on the project (called Arrowsmith) can be found by clicking on http://arrowsmith2.psych.uic.edu/.
Other Information:
Dr. Vetle I. Torvik's homepage: http://arrowsmith2.psych.uic.edu/torvik
Access Vetle's Ph.D. Defense slides as a PDF file (size = 130 KB).
Access Vetle's Ph.D. Dissertation as a PDF file (size = 1.70 MB).
Access Vetle's DTD (Electronic Theses & Dissertations) record of his dissertation at LSU's Library.
E-mail: vtorvik@uic.edu
In 2000 Vetle I. Torvik, while a Ph.D. student of Dr. Triantaphyllou, received a Louisiana Engineering Foundation's Vincent A. Forte Graduate School Fellowship for 2000. This prestigious fellowship was accompanied by $2,000. Only two such awards were available for the entire State of Louisiana. Vetle was selected among many candidates from the State of Louisiana "because of his outstanding accomplishments and his desire to enter the teaching profession in the field of engineering."
Some pictures from Vetle's graduation in May of 2002.
The first picture depicts Vetle and Dr. Triantaphyllou (the two Dr. T's) and the second picture depicts Vetle, Vetle's brother John and his proud father eatingh sushi at Komodo's Restaurant in Baton Rouge, LA. Since both Vetle's brother and father are also Doctors (the first in Electrical Engineering and the second in Cardiology) that picture depicts four (4) Dr. T's at once. It is a rather rare occasion to have so many Dr. T's (3 Dr. Torviks and 1 Dr. Triantaphyllou) all at once!
Photo #A.1, Photo #A.2,
Some pictures when Egemen Yilmaz graduted on May 2000. The place was the Zorbas Greek Restaurant in Baton Rouge, LA.
The depicted people are: Elizabeth, Vetle, a friend, and Egemen.
Photo #B.1, Photo #B.2, Photo #B.3, Photo #B.4
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
3 Published refereed journal papers
1 pending refereed journal papers
1 Refereed proceedings paper
1 Refereed book chapter
4 Conference and other presentations
Graduated with the Ph.D. degree in December of 1999, Ph.D. Committee Chairman.
Ph.D. Dissertation Title:
Classification of Text Decuments Using a Logic Based Approach
ABSTRACT:
The main problem investigated in this dissertation is a follows: Given are two samples of documents each from one of two disjoint collections of documents. The question is how to obtain a set of patterns of text features that make a document in the two samples ( and other unclassified documents) to be classified correctly in one and only one document class. A sample of 2,897 documents from the TIPSTER collection was used to investigate this problem.
This problem was divided into the following four subproblems. The first subproblem consists of identifying the set keywords to describe the documents' content. Computational results of twenty experiments suggested that single-word keywords addressed the main problem effectively.
The second subproblem requires a methodology to construct classification rules to infer the class of unclassified documents. A logical analysis approach called the One Clause At a Time algorithm (OCAT) is used to address this problem. Its accuracy is compared to the one of the Vector Space Model (VSM), a benchmaking methodology in document classification processes. Under identical experimental conditions, some computational results suggests that OCAT algorithm is more accurate that the VSM to solve the main problem.
The third subproblem consists of providing a methodology to construct better rules as more documents become available. This problem has been investigated using the OCAT algorithm under a guided and a random learning approach. Computational results on three samples of 510 documents indicate that the guided learning approach constructs more accurate rules.
In the fourth subproblem an incremental version of the OCAT algorithm is required. The algorithm is needed to speed up the construction of the classification rules. Computational results on three samples of 336 documents each show that: (i) the classification rules become accurate more rapidly, (ii) the CPU times are substantially reduced, and (iii) the rules become more complex as more documents were added to the experiment.
In summary, the results of this research suggest with high confidence that the incremental OCAT algorithm can perform better than the VSM and that it can deliver better and faster results for the classification of large collections of documents.
Current Employment:
In 2001 Dr. Nieto was with Motorola Inc. in Taipei, Taiwan
In 2003 Dr. Nieto was with Motorola Inc. in Nogales, Mexico
Other Information:
E-mail: SalvadorNieto@motorola.com
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
3 Published refereed journal papers
1 Refereed book chapter
2 Conference and other presentations
M.S., May of 2000. Committee Chairman.
M.S. Thesis Title:
A Polynomial Time Heuristic for Mining Association Rules
ABSTRACT:
Mining association rules from databases has attracted great interest because of its potentially very useful results. Taking purchase records of all shoppers into account at once, there are likely to be many sets of items that tend to be purchased together across the group. Presumably, this is the information a store manager could use to make decisions about where to place items in the store so as to increase sales. This information can be expressed in the form of association rules. Given a database, the problem of interest is how to mine association rules out of that very database in an efficient and not costly way. In today's world, the databases involved can be very large. Thus, fast and effective algorithms are needed to mine association rules out of them. Previous methods in this area cause exponential resource consumption and storage capacity. The combinatorial explosion is a natural result of the previous work, because they exhaustively mine all the rules.
This thesis research takes a previously developed approach, called the Randomized Algorithm 1 heuristic, and makes changes to it to mine association rules out of databases in an efficient way. The approach was primarily developed for inferring logical clauses from examples. The new method is called Altered Randomized Algorithm 1 (ARA1). ARA1, by its nature, always produces 100% confidence level rules, which is making the approach a very valuable one.
It is shown via a number of tests that ARA1 is beating the core of all association rule mining techniques, namely Apriori. ARA is also compared to a data mining software called Mineset and it was seen that ARA1 also beats Mineset. The results show that ARA1 is a fresh, successful addition to the field of association rule mining.
Current Employment:
Currently (2001) Egemen is with with GE Capital in Connecticut
Other Information:
E-mail: egemen.yilmaz@gecapital.com
CONGRATULATIONS to Aniruddha S. Deshpande (M.S. in IE, 1995) and Egemen Yilmaz (M.S. in IE, 2000), former students of Dr. Triantaphyllou, who were awarded one Summit Award each for excellent performance in 2001 from GE Capital (click here to see Amiruddha's award and here to see Egemen's award). This is the second highest award in the company. Aniruddha and Egemen have been among the best graduate students Dr. Triantaphyllou ever had.
Some pictures when Egemen Yilmaz graduted on May 2000. The place was the Zorbas Greek Restaurant in Baton Rouge, LA.
The depicted people are: Elizabeth, Vetle, a friend, and Egemen.
Photo #1, Photo #2, Photo #3, Photo #4
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
1 Published refereed journal paper
M.S., May of 1999. Committee Chairman.
M.S. Thesis Title:
A Metaheuristic Approach for the Single Machine Tardiness Problem
ABSTRACT:
For many problems, different heuristics produce the best results under different conditions. Suppose an instance of such a problem is to be solved. Based on the characteristics of the problem instance, a heuristic is chosen. During the solution of the problem instance, the heuristic in use makes a decision that causes an item (node, arc, job, etc.) to be removed from the unsolved problem and to become fixed in the growing partial solution. The remaining items from a smaller version of the same problem with possibly different characteristics. It may be advantageous to switch from one heuristic to another if the characteristics of the unsolved problem change significantly.
A method is herein proposed to create a metaheuristic that monitors the changing characteristics of the problem as each heuristic decision is made, and attempts to employ the best heuristic during each pass. A set of rules guides the decision-making process of the metaheuristic. The single machine tardiness problem was used as a case study of the approach.
Current Employment:
Currently (2001) Peter is with Ernest & Young Consulting in Houston, TX.
E-mail: pjhaynes@yahoo.com
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
1 pending refereed journal paper
1 Conference presentation
M.S., December of 1997. Committee Chairman.
M.S. Thesis Title:
Inference of Boolean Functions from Incomplete Data
ABSTRACT:
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
1 Refereed book chapter
1 Conference presentation
M.S., December of 1996. Committee Chairman.
M.S. Thesis Title:
Estimating Missing Pairwise Comparisons and Which Comparison to Elicit Next
ABSTRACT:
In multi-criteria decision making process, the incomplete AHP is a topic that has long been discussed in the literature. It has many advantages over the complete one, such as saving decision maker's time and effort and become more flexible. However, there is no significant computational experience in using it. This research provides some information and insights that make the incomplete AHP more applicable. Two research problems are investigated in this thesis. Problem 1 is to estimate the missing comparisons in an incomplete comparison matrix. Problem 2 is to choose the most critical comparison to evaluate next. A set of simulation experiments are tested on different combinations of designed parameters (i.e., the estimation method used, the dimension of the comparison matrix, and the percentage of missing comparisons). Then, the rules were analyzed to check the accuracy of the estimation methods and the effectiveness of the guidance rule in selecting the next comparison to evaluate.
Two types of measures of performance are used in the analysis, (i) the top ranked alternative, and (ii) the ranking of the entire set of alternatives. Moreover, there are several alternative measures for the entire ranking, such as the percentage of agreements, the entire correctness, and the sum of squares of the differences (as defined later in this thesis). Also, some related interesting issues (i.e., the computational requirement, the relationship with the consistency of the comparison matrix) are examined. The computational results of this thesis will give an effective guidance to decision makers using multi-criteria decision making methods.
Current Employment:
Currently (2001) Qing is with GE Capital in Connecticut. Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.edu/trianta for the details)
-
1 pending refereed journal papers
1 Refereed book chapter
1 Conference presentation
M.S., August of 1995. Committee Chairman.
M.S. Thesis Title:
Construction of Logical Decision Rules for Pattern Classification from Complete and Incomplete Data
ABSTRACT:
A common problem in many scientific, engineering and business applications is to explain the occurrence of certain desirable or undesirable events based on some definite characteristics or combination of these characteristics. For instance, breast tumors are classified as benign or malignant. The research problem in the above case is to identify patterns among certain biological attributes which can be used to classify tumors as benign or malignant. This research solves the problem of pattern classification from distinct and disjoint membership classes by means of a logic analysis approach. Logic analysis attempts to construct logical decision rules from two or more classes of distinct and disjoint observations. The output of this methodology is a set of logical decision rules as opposed to other procedures such as neural network or linear programming, thus making it much easier and more lucid for the field expert to understand the final outcome. This research deals with three types of interrelated problems.
Problem 1 is how to derive Boolean functions from positive (successful) and negative (unsuccessful) examples. Problem 2 is the use of a guided learning approach for inferring the Boolean logic from positive and negative examples which will minimize the number of examples needed to infer the underlying logic. Problem 3 deals with the inclusion of the incomplete data set along with the positive and negative examples.
An important point to be noted is that while many of the previous methods deal with similar problems in exponential time, the algorithms in this research infer a small number (may not necessarily be the minimum possible) of CNF (conjunctive normal form) rules in polynomial time complexity. The approach assumes that for the same input data, the output is fixed. Computational results indicate that the proposed algorithms would be effective for solving large problems in real world applications.
- In the spring of 1996 Aniruddha won the Second Place
Award of Graduate Research,
Current Employment:
Currently (2001) Aniruddha is with GE Capital in Connecticut.
E-mail: aniruddha.deshpande@gecapital.com
Congratulations to Aniruddha S. Deshpande, a former graduate student of Professor Triantaphyllou (M.S. in 1995), who was just promoted to the position of Chief Risk Officer for GE’s (General Electric’s) subsidiary in Japan!
Thus, Aniruddha will move to Tokyo, Japan, for the next few years. While was at LSU Aniruddha had won a national prestigious research award for his M.S. under Professor Triantaphyllou and later numerous achievement awards for his work at GE. You can learn more about Aniruddha’s work at LSU by visiting the
Date posted: April 6, 2005
CONGRATULATIONS to Aniruddha S. Deshpande (M.S. in IE, 1995) and Egemen Yilmaz (M.S. in IE, 2000), former students of Dr. Triantaphyllou, who were awarded one Summit Award each for excellent performance in 2001 from GE Capital (click here to see Amiruddha's award and here to see Egemen's award). This is the second highest award in the company. Aniruddha and Egemen have been among the best graduate students Dr. Triantaphyllou ever had.
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
2 Published refereed journal papers
1 Refereed proceedings paper
2 Conference and other presentations
M.S., December of 1995. Committee Chairman.
M.S. Thesis Title:
Evaluation of Different Ranking Techniques in Multi-Criteria Decision Making
ABSTRACT:
This research presents an analysis of two ranking approaches that are used in mulit- criteria decision making (MCDM). The ranking approaches examined are the benefit to cost ratio approach and the benefit minus cost approach. The MCMD methods used in this study are the weighted sum model, the analytic hierarcy process, the revised analytic hierarchy processes, and the weighted product model. The two ranking approaches are used by the previous four decision making methods to deal with conflicting criteria such as benefit and cost criteria. The two ranking approaches are seen to give different best alternative selections and also different rankings of the alternatives for the same decision making problem using the same MCDM method. This contradiction is studied in depth for numerous simulated decision problems. The tests are done on simulated decision problems in which random numbers are generated and used as the input data. These data are the performance of the alternatives in term of each one of the decision criteria and the weights of importance of the decision criteria. Problems with different numbers of alternatives and criteria are considered. This study is repeated for each one of the four MCDM methods. This study shows a high contradiction rate in the best alternative selection and in the ranking of the alternatives for the same decision making problem using the same MCDM method, when the two ranking approaches are used. The contradiction in best alternative selection and rankings changes depending on the number of alternatives and the number of criteria in the decision making problem. Finally an effort is given to arrive at a guideline as to what ranking approach gives high rate of contradiction for a particular size decision matrix and based on that when MCDM method to use when dealing with conflicting criteria for different size decision problems.
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
1 pending refereed journal papers
1 Conference and other presentations
M.S., December of 1995. Committee Chairman.
M.S. Thesis Title:
A Study of Some Lot Sizing Techniques in Material Requirement Planning Systems
ABSTRACT:
The performance of a Material Requirement Planning (MRP) system is affected by many factors; one of which is the lot-sizing technique used. The literature review showed that simplicity in the calculations is the main reason for choosing a particular lot-sizing technique. The literature review also showed that benchmarking of lot-sizing techniques is necessary. The goal of this benchmarking is to determine the relative performance versus techniques yielding optimal results. Despite these comparisons, a new technique (the AHD technique) has not been benchmarked versus popular and already accepted lot-sizing techniques (i.e., L4L, FPQ, LUC, and SMH).
The simulation experiments in this thesis compared the performances of these five lot-sizing techniques. These experiments considered: (1) different numbers of items, (2) five demand levels, (3) fixed setup to holding cost ratios, and (4) fixed processing routing. The performance measures for these comparison were: the total cost, and the CPU times, both measures as functions of the reorder interval.
The results were statistically pair-wise compared. They showed that for low demands, the ADH technique always produced lower costs than the other four techniques. However, for medium and larger demands, all techniques performed identically to the L4L technique. The results regarding CPU times showed that the ADH technique always took longer times to find the solution than the other four techniques did. The central conclusion of this study was that the ADH technique can be used for demands relatively low and in the presence of few numbers of items.
First employment:
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
2 Published refereed journal papers
1 Refereed book chapter
2 Conference and other presentations
M.S., May of 1992. Committee Chairman.
M.S. Thesis Title:
A Heuristic For Discrete Search Problems With Positive Switch Costs
ABSTRACT:
An object is hidden in one of n cells according to a known probability distribution. A search policy s is a sequence of cells to be visited and searched in attempt to find the target. The probability of overlooking the target is we search cell i and if the target is in cell i is ai. The cost of searching cell i is ci. The cost of moving, or switching, from cell j to cell i is mji. An optimal policy, s, is one for which the expected cost of finding the target is a minimum. When all mji = 0 the problem has a well known solution. The problem with positive mji is NP-hard and there is not an easy solution. We provide a heuristic to construct a search policy, s', which is good, but is not guaranteed to be optimal. The heuristics is an extension of the optimal solution for the problem with zero switching costs.
First employment:
Instructor at the West Point Military Academy
M.S., January of 1992. Committee Chairman.
M.S. Thesis Title:
Identification of the Critical Criteria When the Analytic Hierarchy Process is Used
ABSTRACT:
Very often data in decision making problems are imprecise and changeable. Therefore, a sensitivity analysis is required. The findings of this paper reveal that when the Analytic Hierarchy Process is used in a decision making problem, sometimes the criterion with the highest weight might be less important than the criterion with the lowest weight.
First employment:
Assistant Professor, Texas Tech University.
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
1 Published refereed journal paper
3 Conference and other presentations
M.S., May of 1992. Committee Chairman.
M.S. Thesis Title:
Inference of a 3-D Object From a Partial 2-D Projection
ABSTRACT:
Object recognition has been an active research field since the early 1979s. It is a prime requirement for many industrial applications, such as motion analysis, expert systems, and diagnostic systems. In industry, object recognition also plays an important role on robotic assembly, inspection of mechanical parts, Flexible Manufacturing Systems (FMSs), and Computer Aided Process Planning (CAPP).
The objective of this thesis is to present a new object recognition methodology. Compared to the matching method used by typical recognition systems, the inference methodology presented here provides a more flexible recognition process. This inference approach can directly infer the framework of an object from a given view without accessing data from a database. Some algorithms are developed to help infer the entire geometry of an object and clarify the ambiguities which are resulted from vision occlusions. An example taken from the literature is inferred in order to demonstrate the inference methodology.
First employment:
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
1 Refereed proceedings paper
1 Conference presentation
M.S., May of 1992. Committee Chairman.
M.S. Thesis Title:
Fuzzy Multi-Attribute Decision-Making
ABSTRACT:
In this thesis we develop five fuzzy decision-making methods and we also evaluate them in terms of two evaluative criteria. An extension of the original Analytic Hierarchy Process (or AHP) was done by Lootsma by using fuzzy triangular numbers to assess the relative importance of a set of alternatives in terms of a set of criteria. Similarly with the fuzzy AHP, five more deterministic decision-making methods are fuzzified in this research. They are the Weighted Sum Model (or WSM), the Weighted Product Model (or WPM), the Revised Analytical Hierarchy Process (or RAHP), the ELECTRE method, and the TOPSIS method. Three scales, the Saaty's scale, the Lootsma's scales, and a New scale are used in this research to quantify the linguistic statements for estimating the fuzzy data. Two evaluative criteria are also developed and used in order to examine the performance of the previous fuzzified decision-making methods in an attempt to find the best method. The findings in this research reveal that the fuzzy Revised Analytic Hierarchy Process outperforms all the other five fuzzy methods.
First employment:
Publications related to this graduate research:
(Please visit Dr. Triantaphyllou's webpage at: http://www.csc.lsu.edu/trianta for the details)
-
1 Published refereed journal paper
1 Conference presentation
M.S., January of 1992. Committee Chairman.
M.S. Thesis Title:
The Problem of Minimizing the Total Question Asking Cost in Horn Clause Systems
ABSTRACT:
Generally defined, experts are good at solving specific types of problems. Their skills usually come from extensive experience, and detailed specialized knowledge of the problems they handle. An expert system is a computer program which captures the human specialist's knowledge, so that it can solve problems expertly.
An expert system consists mainly of two major parts: the knowledge base and the inference engine. A knowledge engineer works closely with the domain experts and captures the experts' knowledge into the knowledge base, which is usually in the form of a set of production rules. Very often these rules are in the form of Horn clauses. The inference engine is a reasoning mechanism, in the form of a computer program, which solves problems by applying a question asking strategy to the knowledge base.
An inference engine applies a question asking strategy on the rules in the knowledge base and reaches a conclusion set by the user of the expert system at the beginning of a consultation season. Usually, if the relative knowledge is insufficient to reach a conclusion, then the expert system has to ask the user questions for additional information. These questions are about the values of decision variables present in the production rules of the knowledge base. Therefore, the number of questions which an expert system asks is a critical issue.
A good question asking strategy of an inference engine should be an efficient strategy, i.e., one which quickly selects the few key questions necessary to reach a conclusion. The more questions the system asks, the more time it takes to reach a conclusion. Also, it is more likely for the user to enter a wrong answer. Therefore, an expert system needs a question asking strategy which can ask as few questions as possible. However, answering a question may involve a cost. In this thesis, a question-asking strategy is developed which attempts to minimize the total cost in reaching a conclusion.
Current Employment:
Currently (2001) Ted works for a major financial company called International Bill Finance Corporation in Taiwan. The URL of this company is: http://www.ibfc.com.tw.
In the fall of 2003 Ted was promoted to the rank of General Manager of the Information Office at IBFC. Congratulations to Ted! Ted was Professor's Triantaphyllou very first graduate student.
E-mail: tedlisa75@yahoo.com.tw or:
E-mail: a384@ibfc.com.tw
Other Information:
Photo #1, Photo #2, Photo #3