Ebook: Fuzzy Systems and Data Mining II
Fuzzy systems and data mining are now an essential part of information technology and data management, with applications affecting every imaginable aspect of our daily lives. This book contains 81 selected papers from those accepted and presented at the 2nd international conference on Fuzzy Systems and Data Mining (FSDM2016), held in Macau, China, in December 2016.
This annual conference focuses on 4 main groups of topics: fuzzy theory, algorithm and system; fuzzy applications; the interdisciplinary field of fuzzy logic and data mining; and data mining, and the event provided a forum where more than 100 qualified, high-level researchers and experts from over 20 countries, including 4 keynote speakers, gathered to create an important platform for researchers and engineers worldwide to engage in academic communication.
All the papers collected here present original ideas, methods and results of general significance supported by clear reasoning and compelling evidence, and as such the book represents a valuable and wide ranging reference resource of interest to all those whose work involves fuzzy systems and data mining.
Fuzzy Systems and Data Mining (FSDM) is an annual international conference devoted to four main groups of topics: a) fuzzy theory, algorithm and system; b) fuzzy application; c) the interdisciplinary field of fuzzy logic and data mining; and d) data mining. Following the great success of FSDM 2015, held in Shanghai, the second edition in the FSDM series was held in Macau China where experts researchers academics participants from the industry were introduced to the latest advances in the field of Fuzzy Sets and Data Mining. Macau was declared a UNESCO World Heritage Site in 2005 by virtue of its cultural importance. The historic centre of Macau is of particular interest because of its mixture of traditional Chinese and Portuguese cultures. Macau has both Cantonese (a variant of Chinese) and Portuguese as official languages.
This volume contains the papers accepted and presented at the 2nd International Conference on Fuzzy Systems and Data Mining (FSDM 2016), held on 11–14 December 2016 in Macau, China. All papers have been carefully reviewed by programme committee members and reflect the breadth and depth of the research topics which fall within the scope of FSDM. From several hundred submissions, 81 of the most promising and FAIA mainstream-relevant contributions have been selected for inclusion in this volume; they present original ideas, methods or results of general significance supported by clear reasoning and compelling evidence.
FSDM 2016 was also a reference conference, and the conference programme included keynote and invited presentations, oral and poster contributions. The event provided a forum where more than 100 qualified and high-level researchers and experts from over 20 countries, including 4 keynote speakers, gathered to create an important platform for researchers and engineers worldwide to engage in academic communication.
I would like to thank all the keynote and invited speakers and authors for the effort they have put into preparing their contributions to the conference. We would also like to take this opportunity to express our gratitude to those people, especially the program committee members and reviewers, who devoted their time to assessing the papers. It is an honour to continue with the publication of these proceedings in the prestigious series Frontiers in Artificial Intelligence and Applications (FAIA) from IOS Press. Our particular thanks also go to J. Breuker, N. Guarino, J.N. Kok, R. López de Mántaras, J. Liu, R. Mizoguchi, M. Musen, S.K. Pal and N. Zhong, the FAIA series editors, for supporting this conference.
Last but not least, I hope that all our participants have enjoyed their stay in Macau and their time at the Macau University of Science and Technology (M.U.S.T.). We hope you had a magnificent experience in both places.
Antonio J. Tallón-Ballesteros
University of Seville, Spain
Issue of deciding interval length, calculations of complicated fuzzy logical relations and hunt of apposite defuzzification process have been an important area of research in fuzzy time series forecasting since its inception. In present study, cumulative probability distribution based computational scheme with discretized of universe is proposed for fuzzy time series forecasting. In this study, cumulative probability distribution decides the length of intervals using characteristic of data distribution and proposed computational algorithm minimizes calculations of complex fuzzy logical relations and search of suitable defuzzification method. To verify the enhancement in forecasting accuracy of developed model, it is applied to the benchmark problem of forecasting historical student enrollments of University of Alabama. Accuracy in forecasted enrollments of developed model is also compared with the other various methods using different error measures. Coefficients of correlation and determination are used to determine the strength between forecasted and actual enrollments.
In this communication the formulation of optimization problems using fuzzy dual parameters and variables is introduced to cope with parametric or implementation uncertainties. It is shown that fuzzy dual programming problems generate finite sets of deterministic optimization problems, allowing to assess the range of the solutions and of the resulting performance at an acceptable computational effort.
In this paper, a sports outcome prediction approach based on sports metric candlestick and fuzzy pattern recognition is proposed. The sports gambling market data are gathered and processed to form the candlestick chart, which has been widely used in financial time series analysis. Unlike the traditional candlestick is composed of the price for financial market analysis, the candlestick for sports metric is determined by the point spread, total point scored, and the gambling shock which measures the bias of gambling line and real total point scored. The fluctuation behaviors of sports outcome are represented by the fuzzification of candlestick for pattern recognition. The decision tree algorithm is applied on the fuzzified candlesticks to find the implicit knowledge rules, and used these rules to forecasting the sports outcome. The National Football League is introduced to our empirical study to verify the effectiveness of forecasting.
Aiming at the problem of high cost and slow equalization speed in traditional circuit, a parallel filling valley equalization circuit based on fuzzy control is proposed in this paper. A fuzzy controller suitable for the circuit is designed. The average voltage, voltage range and the balance electric quantity of the battery pack are modeled by fuzzy model. The fuzzy reasoning and defuzzification is produced to optimize the circuit control logic, which can be adapted to the nonlinearity of the battery pack and the uncertainty of the battery parameters. The simulation and experiment results show that, in the process of charging and discharging, the fuzzy control based parallel filling valley equalization circuit has the advantage of fast and efficient equalization which can improve the use efficiency of the battery pack.
Hesitant fuzzy set (HFS) is one of the most common used techniques for expressing the decision maker's subjective evaluation information. Interval-valued hesitant fuzzy set (IVHFS) is the extension of HFS and can reflect our intuition more objectively. In this paper we focus on the IVHF information aggregation methods based on Bonferroni mean (BM). We proposed the IVHF geometric BM operator (IVHFGBM) and weighted IVHFGBM operators. Some numerical examples for the operators are designed for showing their effectiveness. The desirable properties of weighted IVHFGBM operator are also discussed in detail. These operators can be applied in many areas especially in decision making problems.
Most of the integrated methods of multi-attributes decision making (MADM) used type-1 fuzzy sets to represent uncertainties. Recent theory has suggested that interval type-2 fuzzy sets (IT2 FS) could be used to enhance representation of uncertainties in decision making problems. Differently from the typical integrated MADM methods which directly used type-1 fuzzy sets, this paper proposes an integrating simple additive weighting – technique for order preference similar to ideal solution (SAW-TOPSIS) based on IT2 FS to enhance judgment. The SAW with IT2 FS is used to determine the weight for each criterion, while TOPSIS method with IT2 FS is used to obtain the final ranking for the attributes. A numerical example is used to illustrate the proposed method. The numerical results show that the proposed integrating method is feasible in solving MADM problems under complicated fuzzy environments. In essence, the integrating SAW-TOPSIS is equipped with IT2 FS in contrast to type-1 fuzzy sets for solving MADM problems. The proposed method would make a great impact and significance for the practical implementation. Finally, this paper provides some recommendations for future research directions.
In this paper, we focus on describing the oscillation period and index of fuzzy tensor. The definition of the induced third-order fuzzy tensor is proposed. By using this notion, firstly, the oscillation period and index of fuzzy tensor are obtained on the basis of Power Method with max-min operation. Secondly, we rely on Minimal Strong Component to find the oscillation period of fuzzy tensor. It is a more practical graph theory method that the number of nonzero elements is less than half of the sum of fuzzy tensor elements. Furthermore, numerical results demonstrate the Power Method and the Minimal Strong Component two algorithms for solving the period and index of fuzzy tensor which are effective and promising.
In this paper, we have proposed a mathematical formulation of fuzzy minimum cost flow problem for damageable items transportation. For the imprecise model, capacity, cost, percentage of unit damage of each route have been considered as triangular fuzzy numbers. This problem has been solved by using the k-preference integration method, the area compensation method, and the signed distance method. Finally, to show the validity of the proposed model, a numerical example is provided and solved with Wolfram Mathematica 9.
E-Commerce is a business mode based on internet and information technology. Data mining techniques are widely used in E-Commerce for digging out patterns and retrieving information from large scale noisy datasets. The booming of E-Commerce enables businesses to collect large amount of data which could be analyzed for enhancing revenues. The abundant data collected online is the foundation of big data analysis. How to employ data mining models on strategizing and making business decisions is an important topic in recent years. This paper talks about data mining and its application in E-Commerce. The application of data mining in electronic commerce developed based on data mining technology of electronic commerce system to strengthen the ability of business information analysis, it is concluded that the intrinsic relationship between data and extract the useful information, to provide the expected information of the electronic commerce for the business management personnel, to ensure the effective operation of the electronic commerce. Data mining techniques could be used for automated data analysis, pattern identification, information retrieving, business strategizing as well as providing personalized services.
With the rapid development of the Semantic Web research, the demand for representation and reasoning with uncertain information increases. Despite ontology is capable of modeling the semantic and knowledge in knowledge-based system, classical ontology languages are not appropriate to deal with uncertainty in knowledge, which is inherent in most of the real world application domains. In this paper, we address this issue by extending the power of expression in current ontology language, that is, proposing a Fuzzy Multi Entity Bayesian Networks ontology language which extends the PR-OWL and based on combination of Fuzzy MEBN and ontology, defining and studying its syntax and semantics, and showing representation of domain knowledge by RDF graphs. The proposed language Fuzzy PR-OWL will move beyond the current limitation of modeling the knowledge with fuzzy semantic or fuzzy relation in PR-OWL. By providing a principled means of uncertainty representation and reasoning, Fuzzy PR-OWL can serve for many applications with fuzzy and probability knowledge.
Return voltage method (RVM) is a good method to study aging state of transformer insulation, but it is difficult to accurately assess the insulation aging state by a single characteristic quantity. In this paper, the fuzzy rough sets theory combined with RVM is proposed to assess the oil paper insulation state of transformer and construct the assessment system of oil paper insulation of transformer based on a lot of test data. First, the evaluation index of oil-paper insulation status of transformer is established by return voltage characteristic parameters. Then, fuzzy c-means clustering algorithm is used to obtain the membership function of the transformer test data along with fuzzy partition of characteristics .Moreover, the fuzzy attributes of assessment table of oil paper insulation statue is simplified according to the distinct matrix,and it extracts the evaluation rule of oil paper insulation condition. Finally, the examples in this paper demonstrate that the assessment system is effective and feasible, which provides a new idea for the assessment of transformer oil-paper insulation state. The research has practical value in application of engineering
The finite-time stabilization problem for nonlinear networked systems has been considered. T-S approach has been used to model the controlled nonlinear systems. By using the Lyapunov functional method, a finite-time stabilization sufficient condition has been given. Then, a state feedback fuzzy controller has been designed to make the closed networked control systems finite-time stable. Finally, the proposed design method has been used into the temperature control system for polymerization reactor.
Fuzzy multiple attribute decision making (FMADM) is an efficient way to solve complex systems, and has wide, practical application. This paper studies the FMADM of the trapezoidal fuzzy number. In order to achieve desirable decision making, the similarity measures between two trapezoidal fuzzy numbers is defined, which is based on a new method for ranking fuzzy numbers. A new algorithm is proposed to remove surplus attributes. This algorithm is based on rough sets and the technique for order of preference by similarity to ideal solution (TOPSIS) method; Finally, an example is examined to demonstrate the model's use in practical problems.
Stock market investing is an inherently risky and imprecise activity, requiring complex decision making under uncertainty. This paper proposes a method that applies fuzzy rule-based inference to rank stocks based on price momentum and market capitalization. Experiments performed on Thai stock market data showed that high-momentum stocks significantly outperformed the market index benchmark, and that stocks of companies with small market capitalization performed better than larger ones. Fuzzy rule-based inference was applied to combine both the momentum factor and the market capitalization factor, with different sets of rules for different prevailing market conditions. The result produced a higher investment return than using either momentum or market capitalization alone.
Aiming at control method of 2-DOF joint robot, the 3D robot model is established in ADAMS firstly, and then dynamic equation of the robot is deduced by using the obtained parameters. And dynamic model is combined with control system model in MATLAB/Simulink by the ADAMS/Control module and is established coordinated simulation system. In order to eliminate the effect of the modeling error and uncertainty signal, a sliding-mode control is proposed. In this method, a linear sliding surface is used to ensure the system to reach equilibrium with the sliding surface in finite time; and fuzzy control is used to compensate for the modeling error and uncertainty signal. Equivalent control law and switching control law are derived by using Lyapunov stability criterion and exponential reaching law. Fuzzy control law and membership function are set up by using fuzzy control rules. Through online adaptive learning of fuzzy, buffeting is weakened. Simulation result shows that the control method is effective.
In this paper, by combining hesitant fuzzy set with bipolar-valued fuzzy set, the concept of hesitant bipolar value fuzzy set is introduced, and the hesitant bipolar fuzzy group decision making method based on TOPSIS is proposed. Our study firstly integrates fuzziness, hesitation and incompatible bipolarity in multiple criteria decision making method. An illustrative case of chemical project evaluation also demonstrates the feasibility, validity, and necessity of our proposed method.
In this paper, using chance constrained programming formulation, a new chance constrained twin support vector machine (CC-TWSVM) is proposed. This paper studies twin support vector machine classification when data points are uncertain with measurement statistically noise. With some properties known for the distribution, the CC-TWSVM model aims to ensure the small probability of error classification for the uncertain data. We also provide equivalent second-order cone programming (SOCP) model of the CC-TWSVM model by the properties of moment information of uncertain data. The dual problem of SOCP model is introduced and the optimal value of the CC-TWSVM model can be solved directly. In addition, we also show the performance of CC-TWSVM model in artificial data and real data by numerical experiments.
This paper deals with non-algebraic binary relational semantics, called here set-theoretic Kripke-style semantics, for monoidal t-norm (based) logics. For this, we first introduce the system MTL (Monoidal t-norm logic) and some of its prominent axiomatic extensions, and then their corresponding Kripke-style semantics. Next, we provide set-theoretic completeness results for them.
Handling dynamic aspect of databases and multiple support threshold requirements of items are two important challenges of frequent itemset mining algorithms. Existing dynamic itemset mining algorithms are devised for single support threshold whereas multiple support threshold algorithms assume that the databases are static. This paper focuses on dynamic update problem of frequent itemsets under MIS (Multiple Item Support) thresholds and introduces Dynamic MIS algorithm. It is i) tree based and scans the database once, ii) considers multiple support thresholds, and iii) handles increments of additions, additions with new items and deletions. Proposed algorithm is compared to CFP-Growth++ and findings are; in dynamic database 1) Dynamic MIS performs better than CFP-Growth++ since it runs only on increments and 2) Dynamic MIS can achieve speed-up up to 56 times against CFP-Growth++.
In this study, two major applications are introduced to develop advanced deep learning methods for credit card data analysis. Credit card information is contained in two data sets; credit approval dataset and card transaction dataset. The credit card dataset has two problems. One problem is using credit card approval dataset, it is necessary to combine multiple models, each referring to a different clustered group of users. The other problem is using card transaction dataset, since the actual unauthorized credit card use is very small, these imprecise solutions do not allow the appropriate detection of fraud. To solve these problems, we proposed deep learning algorithm to apply credit card dataset. The proposed methods are validated using benchmark experiments with other machine learnings. To evaluate our proposed method, we use two credit card datasets, credit approval dataset by UCI machine learning repository and credit transaction dataset constructed by random. The experiments confirm that deep learning exhibits comparable accuracy to the Gaussian kernel support vector machine (SVM). The proposed methods are also validated using large scale transaction dataset. Moreover, we apply our proposed method for the time-series benchmark dataset. Deep learning parameter adjustment is difficult. By optimizing the parameters, it is possible to increase the learning accuracy.
Uncertain data is the data accompanied with probability, which makes the frequent itemset mining have more challenges. Given the data size n, computing the probabilistic support needs O(n(logn)2) time complexity and O(n) space complexity. This paper focuses on the problem of mining probabilistic frequent itemsets over uncertain databases and proposed PFIMSample algorithm. We employ the Chebyshev inequation to estimate the frequency of the items, which decreases certain computing from O(n(logn)2) to O(n). In addition, we propose the sampling technique to improve the performance. Our extensive experimental results show that our algorithm can achieve a significantly improved runtime cost and memory cost with high accuracy.
In data center networks, energy consumption accounts for a considerably large slice of operational expenses. Many energy saving strategies have been proposed, most of them follow the point of bandwidth or throughput to complete the design of energy saving model. This paper provides a new perspective of energy saving in data center networks, which basic idea is to ensure the higher priority traffics have the shorter routes. Combine the bandwidth constraints with the aim of energy saving, and keep the balance between energy consumption and traffic priority demand. Simulations show that our routing algorithm can effectively reduce the transmission delay of the higher priority traffics and reduce the power consumption of data center networks.