Özge Uncu

A novel feature selection approach: Combining feature wrappers and filters

Abstract Feature selection is one of the most important issues in the research fields such as system modelling, data mining and pattern recognition. In this study, a new feature selection algorithm that combines feature wrapper and feature filter approaches is proposed in order to identify the significant input variables in systems with continuous domains. The proposed method utilizes functional dependency concept, correlation coefficients and K -nearest neighbourhood (KNN) method to implement the feature filter and feature wrappers. Four feature selection methods independently select the significant input variables and the input variable combination, which yields best result with respect to their corresponding evaluation function, is selected as the winner. This is similar to the basic information fusion notion of integrating the information collected from different sources. All of the four feature selection methods are performed in two stages: (i) pre-selection, (ii) selection. Two of the four feature selection methods utilize KNN method for evaluating the candidates. These two methods use sequential forward and sequential backward search mechanism, respectively, in pre-selection stage. Whereas, the third feature selection method uses correlation coefficients in the pre-selection stage. It is common to have outliers and noise in real-life data. In order to make the proposed feature selection algorithm noise and outlier resistant, approximate functional dependencies are used by utilizing membership values that inherently cope with uncertainty in the data. Thus, the fourth feature selection method makes use of approximate functional dependencies to evaluate candidates in pre-selection stage. All of these four methods apply KNN method with exhaustive search strategy in order to find the most suitable input variable combination with respect to a performance measure.

GRIDBSCAN: GRId Density-Based Spatial Clustering of Applications with Noise

Clustering is one of the basic data mining tasks that can be used to extract hidden information from data in the absence of target classes. One of the most well-known density based clustering algorithms for processing spatial data is Density-Based Spatial Clustering of Application with Noise (DBSCAN) that uses learning parameters epsiv and minPts to define the density that will be sought in the data set while forming the clusters. The major drawbacks of the DBSCAN algorithm are its sensitivity to user input required to execute the algorithm, inability to recognize clusters with different densities, and computational complexity. In this study, we propose a three-level clustering method to address the second issue. The first level selects appropriate grids so that the density is homogeneous in each grid. The second stage merges cells with similar densities and identifies the most suitable values of epsiv and minPts in each grid that remain after merging. The third step of the proposed method executes the DBSCAN method with these identified parameters in the dataset. The proposed method is tested in three artificial benchmark data sets to demonstrate that the clusters are correctly identified.

A new fuzzy inference approach based on Mamdani inference using discrete type 2 fuzzy sets

Fuzzy system modeling (FSM) is one of the most prominent system modeling tools in analyzing the data in the presence of uncertainty. Linguistic fuzzy rulebase (LFR) structure, in which both the antecedent and consequent variables are represented by fuzzy sets, is the most well known fuzzy rulebase structure in the literature. The proposed FSM method identifies LFR system model by executing fuzzy C-Means (FCM) clustering method. One of the sources of uncertainty in system modeling is the uncertainty in selecting learning parameters. In order to capture this uncertainty in a more realistic way, the antecedent and consequent variables are represented by using type 2 fuzzy sets that are constructed by executing FCM method with different level of fuzziness, m, values. The proposed system modeling approach is applied on a well-known benchmark data set where the goal is to predict the price of a stock. After comparing the results with the ones obtained with other system modeling tools, it can be claimed successful results are achieved.

An intelligent decision-support system for rough mills

A rough mill is where wood components are cut from lumber to produce wooden doors and windows. Because lumber is a natural material it contains various types of defects (e.g. knots and splits) but their distribution is not known in advance. Furthermore, only the approximate content and dimension of each board in a load are known ahead of time. Thus, producing required components from different types of lumber in a rough mill is quite a complex challenge. Several operations that occur in the rough mill are closely related and are analysed in this paper. An Intelligent Decision-Support System (IDSS) is proposed to improve the overall performance of the rough mill by presenting recommendations to operators. A rough mill simulator helps operators view the effect of these recommendations before implementing them on the production floor. Two key challenges for rough mills are identified, namely: selection of appropriate jags and cut-list scheduling and alternative solutions are proposed.

Framework for Developing Distributed Systems in a Peer-to-Peer Environment

Decentralized software architectures are essential to fully realize the potential of distributed networks. However, developers are burdened with the need to create a communications infrastructure for a distributed system along with their applications. The use of a generic communication interface greatly simplifies the process, thereby allowing developers to focus on applications. This paper describes the design and functionality of such an interface.

Simulation and decision support models for rough mills: a multi-agent perspective

A rough mill is a manufacturing facility where loads of lumber (jags) are processed into specific size components. In this paper, we describe a multi-agent system that simulates the operations of the ripsaw, conveyor and chopsaw, and provides the user with recommended decisions for selecting jags and cut-lists (specific-size components for cutting). Using the graphical user interface, the user can run several scenarios, and test the recommended decisions through simulation. The operator can then make an informed decision on the rough mill floor. A discrete event based simulation model provides functionality for the agent-based model. The Java Agent Development Framework (JADE) was used to develop the agent system. The new system was validated by comparing results obtained using a centralized simulator.

Identifying Aggregation Weights of Decision Criteria: Application of Fuzzy Systems to Wood Product Manufacturing

A rough mill converts lumber into components with prescribed dimensions. The manufacturing process begins with lumber being transferred from the warehouse to the rough mill where its length and width are determined by a scanner. Then, each board is cut longitudinally by the rip saw to produce strips of prescribed widths. Next, the strips are conveyed to another scanner where defects are detected. Finally, the chop saw cuts the strips into pieces with lengths based on the defect information and component requirements. These pieces are conveyed to pneumatic kickers which sort them into bins. The rough mill layout and process of a major Canadian window manufacturer is used throughout this study. The rough mill layout and process, as described by Kotak, et al. [11], is shown in Fig. 1. The operator receives an order in which due dates and required quantities of components with specific dimensions and qualities are listed. Since there is a limited number of sorting bins (which is much less than the number of components in the order), the operator must select a subset of the order, called cut list, and assign it to the kickers. After selecting the cut list, the loads of lumber (called jags) that will be used to produce the components in the cut list must be selected by the operator. The operator also determines the arbor configuration and priority of the rip saw. Methods for kicker assignment have been reported by Siu, et al. [17] and jag selection has been investigated by Wang, et al. [22], [23]. A discrete event simulation model was developed by Kotak, et al. [11] to simulate the daily operation of a rough mill for specified jags, cut lists, ripsaw configuration, and orders. Wang, et al. [22] proposed a two-step method to select the most suitable jag for a given cut list. The first step selects the best jag type based on the proximity of the length distribution of the cut list and the historical length distribution of the

Jag sequencing in rough mill operations

Raw material cost is one of the major contributors to the overall cost in rough mill operations. The challenge is to choose raw materials that can fulfil a given order in a reasonable time. However, the objective of minimizing raw material cost conflicts with the objective of minimizing the processing time. This study investigates the use of a local search mechanism to find the best jag sequence for a given order. Simulation is used to evaluate the performance of each jag sequence candidate with respect to the objective function. Since the proposed method is intended for real-time production, beam search is utilized. Numerical results for a sample order list show 22% cost reduction.

A comparative analysis of fuzzy system modelling approaches: a case in mining medical diagnostic rules

Fuzzy system modeling approximates highly nonlinear systems by means of fuzzy if-then rules In the literature, different approaches are proposed for mining fuzzy if-then rules from historical data These approaches usually utilize fuzzy clustering in structure identification phase In this research, we are going to analyze three possible approaches from the literature and try to compare their performances in a medical diagnosis classification problem, namely Aachen Aphasia Test Given the fact that the comparison is conducted on a single data set; the conclusions are by no means inclusive However, we believe that the results might provide some valuable insights.

Image segmentation using joint clustering analysis of attribute data and relationship data

Attributes of an object contain its fundamental properties. Attribute data is the main source of clustering information. Although relationship data is an extrinsic property of objects and is at least as important as attribute data, most clustering methods process only one type of characteristic data. However, attribute and relationship data must be analyzed together for applications such as market segmentation, social network segmentation, and image segmentation. In this study we describe a new algorithm that combines attribute and relationship data for joint clustering analysis. An experimental evaluation demonstrates the usefulness and accuracy of the proposed algorithm when applied to image segmentation.