An ensemble technique for multi class imbalanced problem using probabilistic neural networks
- Authors: Chandrasekara, N. , Tilakaratne, Chandima , Mammadov, Musa
- Date: 2018
- Type: Text , Journal article
- Relation: Advances and Applications in Statistics Vol. 53, no. 6 (2018), p. 647-658
- Full Text: false
- Reviewed:
- Description: The class imbalanced problem is one of the major difficulties encountered by many researchers when using classification tools. Multi class problems are especially severe in this regard. The main objective of this study is to propose a suitable technique to handle multi class imbalanced problem. Probabilistic neural network (PNN) is used as the classification tool and the directional prediction of Australian, United States and Srilankan stock market indices is considered as the application. We propose an ensemble technique to handle multi class imbalanced problem that is called multi class undersampling based bagging (MCUB) technique. This is a new initiative that has not been considered in the literature to handle multi class imbalanced problem by employing PNN. The results obtained demonstrate that the proposed MCUB technique is capable of handling multi class imbalanced problem. Therefore, the PNN with the proposed ensemble technique can be used effectively in data classification. As a further study, other classification tools can be used to investigate the performance of the proposed MCUB technique in solving class imbalanced problems.
Target learning : A novel framework to mine significant dependencies for unlabeled data
- Authors: Wang, Limin , Chen, Shenglei , Mammadov, Musa
- Date: 2018
- Type: Text , Conference proceedings
- Relation: 22nd Pacific-Asia Conference on Advances in Knowledge Discovery and Data Mining, PAKDD 2018; Melbourne, Australia; 3rd-6th June 2018; published in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) Vol. 10937 LNAI, p. 106-117
- Full Text: false
- Reviewed:
- Description: To mine significant dependencies among predictiveattributes, much work has been carried out to learn Bayesian netwrok classifiers (BNC T s) from labeled training data set T. However, if BNC T does not capture the “right” dependencies that would be most relevant to unlabeled testing instance, that will result in performance degradation. To address this issue we propose a novel framework, called target learning, that takes each unlabeled testing instance as a target and builds an “unstable” Bayesian model BNC P for it. To make BNC P and BNC T complementary to each other and work efficiently in combination, the same learning strategy is applied to build them. Experimental comparison on 32 large data sets from UCI machine learning repository shows that, for BNCs with different degrees of dependence target learning always helps improve the generalization performance with minimal additional computation.
- Description: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Optimality conditions via weak subdifferentials in reflexive Banach spaces
- Authors: Hassani, Sara , Mammadov, Musa , Jamshidi, Mina
- Date: 2017
- Type: Text , Journal article
- Relation: Turkish Journal of Mathematics Vol. 41, no. 1 (2017), p. 1-8
- Full Text:
- Reviewed:
- Description: In this paper the relation between the weak subdifferentials and the directional derivatives, as well as optimality conditions for nonconvex optimization problems in reflexive Banach spaces, are investigated. It partly generalizes several related results obtained for finite dimensional spaces. © Tübitak.
A generalization of a theorem of Arrow, Barankin and Blackwell to a nonconvex case
- Authors: Kasimbeyli, Nergiz , Kasimbeyli, Refail , Mammadov, Musa
- Date: 2016
- Type: Text , Journal article
- Relation: Optimization Vol. 65, no. 5 (May 2016), p. 937-945
- Full Text:
- Reviewed:
- Description: The paper presents a generalization of a known density theorem of Arrow, Barankin, and Blackwell for properly efficient points defined as support points of sets with respect to monotonically increasing sublinear functions. This result is shown to hold for nonconvex sets of a partially ordered reflexive Banach space.
A new reliability analysis method based on the conjugate gradient direction
- Authors: Ezzati, Ghasem , Mammadov, Musa , Kulkarni, Siddhivinayak
- Date: 2015
- Type: Text , Journal article
- Relation: Structural and Multidisciplinary Optimization Vol. 51, no. 1 (2015), p. 89-98
- Full Text: false
- Reviewed:
- Description: Reliability-based design optimization (RBDO) is an important area in structural optimization. A principal step of the RBDO process is to solve a reliability analysis problem. This problem has been considered in inner loop of double-loop RBDO approaches. Although many algorithms have been developed for solving this problem, there are still some challenges. Existing algorithms do not have good convergence rates and often diverge. There is a need to develop more efficient and stable algorithms that can be used for evaluating all performance functions sufficiently. In this paper, a new method, called “Conjugate Gradient Analysis (CGA) Method”, is proposed to apply in the reliability analysis problems. This method is based on the conjugate gradient method. Some mathematical problems are provided in order to demonstrate the advantages of the proposed method compared with the existing methods. © 2014, Springer-Verlag Berlin Heidelberg.
Predicting and controlling the dynamics of infectious diseases
- Authors: Evans, Robin , Mammadov, Musa
- Date: 2015
- Type: Text , Conference proceedings
- Relation: 54th IEEE Conference on Decision and Control, CDC 2015; Osaka, Japan; 15th-18th December 2015; Published in Proceedings of the IEEE Conference on Decision and Control; p. 5378-5383
- Full Text:
- Description: This paper introduces a new optimal control model to describe and control the dynamics of infectious diseases. In the present model, the average time to isolation (i.e. hospitalization) of infectious population is the main time-dependent parameter that defines the spread of infection. All the preventive measures aim to decrease the average time to isolation under given constraints. The model suggested allows one to generate a small number of possible future scenarios and to determine corresponding trajectories of infected population in different regions. Then, this information is used to find an optimal distribution of bed capabilities across countries/regions according to each scenario. © 2015 IEEE.
A hybrid wrapper-filter approach to detect the source(s) of out-of-control signals in multivariate manufacturing process
- Authors: Huda, Shamsul , Abdollahian, Mali , Mammadov, Musa , Yearwood, John , Ahmed, Shafiq , Sultan, Ibrahim
- Date: 2014
- Type: Text , Journal article
- Relation: European Journal of Operational Research Vol. 237, no. 3 (2014), p. 857-870
- Full Text: false
- Reviewed:
- Description: With modern data-Acquisition equipment and on-line computers used during production, it is now common to monitor several correlated quality characteristics simultaneously in multivariate processes. Multivariate control charts (MCC) are important tools for monitoring multivariate processes. One difficulty encountered with multivariate control charts is the identification of the variable or group of variables that cause an out-of-control signal. Expert knowledge either in combination with wrapper-based supervised classifier or a pre-filter with wrapper are the standard approaches to detect the sources of out-of-control signal. However gathering expert knowledge in source identification is costly and may introduce human error. Individual univariate control charts (UCC) and decomposition of T2 statistics are also used in many cases simultaneously to identify the sources, but these either ignore the correlations between the sources or may take more time with the increase of dimensions. The aim of this paper is to develop a source identification approach that does not need any expert-knowledge and can detect out-of-control signal in less computational complexity. We propose, a hybrid wrapper-filter based source identification approach that hybridizes a Mutual Information (MI) based Maximum Relevance (MR) filter ranking heuristic with an Artificial Neural Network (ANN) based wrapper. The Artificial Neural Network Input Gain Measurement Approximation (ANNIGMA) has been combined with MR (MR-ANNIGMA) to utilize the knowledge about the intrinsic pattern of the quality characteristics computed by the filter for directing the wrapper search process. To compute optimal ANNIGMA score, we also propose a Global MR-ANNIGMA using non-functional relationship between variables which is independent of the derivative of the objective function and has a potential to overcome the local optimization problem of ANN training. The novelty of the proposed approaches is that they combine the advantages of both filter and wrapper approaches and do not require any expert knowledge about the sources of the out-of-control signals. Heuristic score based subset generation process also reduces the search space into polynomial growth which in turns reduces computational time. The proposed approaches were tested by exhaustive experiments using both simulated and real manufacturing data and compared to existing methods including independent filter, wrapper and Multivariate EWMA (MEWMA) methods. The results indicate that the proposed approaches can identify the sources of out-of-control signals more accurately than existing approaches. © 2014 Elsevier B.V. All rights reserved.
A new loss function for robust classification
- Authors: Zhao, Lei , Mammadov, Musa , Yearwood, John
- Date: 2014
- Type: Text , Journal article
- Relation: Intelligent Data Analysis Vol. 18, no. 4 (2014), p. 697-715
- Full Text: false
- Reviewed:
- Description: Loss function plays an important role in data classification. Manyloss functions have been proposed and applied to differentclassification problems. This paper proposes a new so called thesmoothed 0-1 loss function, that could be considered as anapproximation of the classical 0-1 loss function. Due to thenon-convexity property of the proposed loss function, globaloptimization methods are required to solve the correspondingoptimization problems. Together with the proposed loss function, wecompare the performance of several existing loss functions in theclassification of noisy data sets. In this comparison, differentoptimization problems are considered in regards to the convexity andsmoothness of different loss functions. The experimental resultsshow that the proposed smoothed 0-1 loss function works better ondata sets with noisy labels, noisy features, and outliers. © 2014 - IOS Press and the authors. All rights reserved.
Sigma supporting cone and optimality conditions in non-convex problems
- Authors: Hassani, Sara , Mammadov, Musa
- Date: 2014
- Type: Text , Journal article
- Relation: Far East Journal of Mathematical Sciences Vol. 91, no. 2 (2014), p. 169-190
- Full Text: false
- Reviewed:
- Description: In this paper, a new supporting function for characterizing non-convex sets is introduced. The notions of σ-supporting cone and maximal conic gap are proposed and some properties are investigated. By applying these new notions, we establish the optimality conditions considered in [7] for a broader class of finite dimensional normed spaces in terms of weak subdifferentials.
Structure learning of Bayesian Networks using global optimization with applications in data classification
- Authors: Taheri, Sona , Mammadov, Musa
- Date: 2014
- Type: Text , Journal article
- Relation: Optimization Letters Vol. 9, no. 5 (2014), p. 931-948
- Full Text:
- Reviewed:
- Description: Bayesian Networks are increasingly popular methods of modeling uncertainty in artificial intelligence and machine learning. A Bayesian Network consists of a directed acyclic graph in which each node represents a variable and each arc represents probabilistic dependency between two variables. Constructing a Bayesian Network from data is a learning process that consists of two steps: learning structure and learning parameter. Learning a network structure from data is the most difficult task in this process. This paper presents a new algorithm for constructing an optimal structure for Bayesian Networks based on optimization. The algorithm has two major parts. First, we define an optimization model to find the better network graphs. Then, we apply an optimization approach for removing possible cycles from the directed graphs obtained in the first part which is the first of its kind in the literature. The main advantage of the proposed method is that the maximal number of parents for variables is not fixed a priory and it is defined during the optimization procedure. It also considers all networks including cyclic ones and then choose a best structure by applying a global optimization method. To show the efficiency of the algorithm, several closely related algorithms including unrestricted dependency Bayesian Network algorithm, as well as, benchmarks algorithms SVM and C4.5 are employed for comparison. We apply these algorithms on data classification; data sets are taken from the UCI machine learning repository and the LIBSVM. © 2014, Springer-Verlag Berlin Heidelberg.
Turnpike theorem for an infinite horizon optimal control problem with time delay
- Authors: Mammadov, Musa
- Date: 2014
- Type: Text , Journal article
- Relation: SIAM Journal on Control and Optimization Vol. 52, no. 1 (2014), p. 420-438
- Full Text:
- Reviewed:
- Description: An optimal control problem for systems described by a special class of nonlinear differential equations with time delay is considered. The cost functional adopted could be considered as an analogue of the terminal functional defined over an infinite time horizon. The existence of optimal solutions as well as the asymptotic stability of optimal trajectories (that is, the turnpike property) are established under some quite mild restrictions on the nonlinearities of the functions involved in the description of the problem. Such mild restrictions on the nonlinearities allowed us to apply these results to a blood cell production model. © 2014 Society for Industrial and Applied Mathematics.
A new auxiliary function method for general constrained global optimization
- Authors: Wu, Zhiyou , Bai, Fusheng , Yang, Yongjian , Mammadov, Musa
- Date: 2013
- Type: Text , Journal article
- Relation: Optimization Vol. 62, no. 2 (2013), p. 193-210
- Full Text:
- Reviewed:
- Description: In this article, we first propose a method to obtain an approximate feasible point for general constrained global optimization problems (with both inequality and equality constraints). Then we propose an auxiliary function method to obtain a global minimizer or an approximate global minimizer with a required precision for general global optimization problems by locally solving some unconstrained programming problems. Some numerical examples are reported to demonstrate the efficiency of the present optimization method. © 2013 Taylor & Francis.
- Description: 2003011103
Attribute weighted Naive Bayes classifier using a local optimization
- Authors: Taheri, Sona , Yearwood, John , Mammadov, Musa , Seifollahi, Sattar
- Date: 2013
- Type: Text , Journal article
- Relation: Neural Computing & Applications Vol.24, no.5 (2013), p.995-1002
- Full Text:
- Reviewed:
- Description: The Naive Bayes classifier is a popular classification technique for data mining and machine learning. It has been shown to be very effective on a variety of data classification problems. However, the strong assumption that all attributes are conditionally independent given the class is often violated in real-world applications. Numerous methods have been proposed in order to improve the performance of the Naive Bayes classifier by alleviating the attribute independence assumption. However, violation of the independence assumption can increase the expected error. Another alternative is assigning the weights for attributes. In this paper, we propose a novel attribute weighted Naive Bayes classifier by considering weights to the conditional probabilities. An objective function is modeled and taken into account, which is based on the structure of the Naive Bayes classifier and the attribute weights. The optimal weights are determined by a local optimization method using the quasisecant method. In the proposed approach, the Naive Bayes classifier is taken as a starting point. We report the results of numerical experiments on several real-world data sets in binary classification, which show the efficiency of the proposed method.
Capped K-NN Editing in definition lacking environments
- Authors: Stranieri, Andrew , Yatsko, Andrew , Golden, Isaac , Mammadov, Musa , Bagirov, Adil
- Date: 2013
- Type: Text , Journal article
- Relation: Journal of Pattern Recognition Research Vol. 8, no. 1 (2013), p. 39-58
- Full Text: false
- Reviewed:
- Description: While any input may be contributing, imprecise specification of class of data subdivided into classes identifies as rather common a source of noise. The misrepresentation may be characteristic of the data or be caused by forcing of a regression problem into the classification type. Consideration is given to examples of this nature, and an alternative is proposed. In the main part, the approach is based on a well-known technique of data treatment for noise using k-NN. The paper advances an editing technique designed around idea of variable number of authenticating instances. Test runs performed on publicly available and proprietary data demonstrate high retention ability of the new procedure without loss of classification accuracy. Noise reduction methods in a broader classification context are extensively surveyed.
Globally convergent algorithms for solving unconstrained optimization problems
- Authors: Taheri, Sona , Mammadov, Musa , Seifollahi, Sattar
- Date: 2013
- Type: Text , Journal article
- Relation: Optimization Vol. , no. (2013), p. 1-15
- Full Text:
- Reviewed:
- Description: New algorithms for solving unconstrained optimization problems are presented based on the idea of combining two types of descent directions: the direction of anti-gradient and either the Newton or quasi-Newton directions. The use of latter directions allows one to improve the convergence rate. Global and superlinear convergence properties of these algorithms are established. Numerical experiments using some unconstrained test problems are reported. Also, the proposed algorithms are compared with some existing similar methods using results of experiments. This comparison demonstrates the efficiency of the proposed combined methods.
Learning the naive bayes classifier with optimization models
- Authors: Taheri, Sona , Mammadov, Musa
- Date: 2013
- Type: Text , Journal article
- Relation: International Journal of Applied Mathematics and Computer Science Vol. 23, no. 4 (2013), p. 787-795
- Full Text:
- Reviewed:
- Description: Naive Bayes is among the simplest probabilistic classifiers. It often performs surprisingly well in many real world applications, despite the strong assumption that all features are conditionally independent given the class. In the learning process of this classifier with the known structure, class probabilities and conditional probabilities are calculated using training data, and then values of these probabilities are used to classify new observations. In this paper, we introduce three novel optimization models for the naive Bayes classifier where both class probabilities and conditional probabilities are considered as variables. The values of these variables are found by solving the corresponding optimization problems. Numerical experiments are conducted on several real world binary classification data sets, where continuous features are discretized by applying three different methods. The performances of these models are compared with the naive Bayes classifier, tree augmented naive Bayes, the SVM, C4.5 and the nearest neighbor classifier. The obtained results demonstrate that the proposed models can significantly improve the performance of the naive Bayes classifier, yet at the same time maintain its simple structure.
Preface: Special issue of JOGO MEC EurOPT 2010-Izmir
- Authors: Kasimbeyli, Refail , Mammadov, Musa , Dincer, Cemali
- Date: 2013
- Type: Text , Journal article
- Relation: Journal of Global Optimization Vol. 56, no. 2 (June 2013), p. 217-218
- Full Text: false
- Reviewed:
- Description: C1
A new method for solving linear ill-posed problems
- Authors: Zhang, Jianjun , Mammadov, Musa
- Date: 2012
- Type: Text , Journal article
- Relation: Applied Mathematics and Computation Vol. 218, no. 20 (2012), p.10180-10187
- Full Text:
- Reviewed:
- Description: In this paper, we propose a new method for solving large-scale ill-posed problems. This method is based on the Karush-Kuhn-Tucker conditions, Fisher-Burmeister function and the discrepancy principle. The main difference from the majority of existing methods for solving ill-posed problems is that, we do not need to choose a regularization parameter in advance. Experimental results show that the proposed method is effective and promising for many practical problems. © 2012.
Application of optimisation-based data mining techniques to medical data sets: A comparative analysis
- Authors: Dzalilov, Zari , Bagirov, Adil , Mammadov, Musa
- Date: 2012
- Type: Text , Conference paper
- Relation: IMMM 2102: The Second International Conference on Advances in Information Mining and Management p. 41-46
- Full Text: false
- Reviewed:
- Description: Abstract - Computational methods have become an important tool in the analysis of medical data sets. In this paper, we apply three optimisation-based data mining methods to the following data sets: (i) a cystic fibrosis data set and (ii) a tobacco control data set. Three algorithms used in the analysis of these data sets include: the modified linear least square fit, an optimization based heuristic algorithm for feature selection and an optimization based clustering algorithm. All these methods explore the relationship between features and classes, with the aim of determining contribution of specific features to the class outcome. However, the three algorithms are based on completely different approaches. We apply these methods to solve feature selection and classification problems. We also present comparative analysis of the algorithms using computational results. Results obtained confirm that these algorithms may be effectively applied to the analysis of other (bio)medical data sets
Global stabilization in nonlinear discrete systems with time-delay
- Authors: Ivanov, Anatoli , Mammadov, Musa , Trofimchuk, Sergei
- Date: 2012
- Type: Text , Journal article
- Relation: Journal of Global Optimization Vol.56, no. 2 (2012), p. 1-13
- Full Text: false
- Reviewed:
- Description: A class of scalar nonlinear difference equations with delay is considered. Sufficient conditions for the global asymptotic stability of a unique equilibrium are given. Applications in economics and other fields lead to consideration of associated optimal control problems. An optimal control problem of maximizing a consumption functional is stated. The existence of optimal solutions is established and their stability (the turnpike property) is proved. © 2012 Springer Science+Business Media, LLC.