Kemal Tutuncu

A fuzzy clustering approach for finding similar documents using a novel similarity measure

Searching for similar documents has a crucial role in document management. This paper aims for developing a fast and high quality method of searching similar documents based on fuzzy clustering in large document collections. In order to perform these requirements, a two layers structure is proposed. Formerly, finding the similarity in documents is based on the strategy that uses word-by-word comparison. The proposed method in this study uses two layers structure and lets the documents pass through it to find the similarities. In this system, predefined fuzzy clusters are used to extract feature vectors of related documents for finding similar documents of them. Similarity measure is estimated based on these vectors. To do this, a distance based similarity measure is proposed. It has been seen in empirical results that the proposed system uses new similarity measure and has better performance compared with conventional similarity measurement systems.

A new approach on search for similar documents with multiple categories using fuzzy clustering

Searching for similar document has an important role in text mining and document management. In whether similar document search or in other text mining applications generally document classification is focused and class or category that the documents belong to is tried to be determined. The aim of the present study is the investigation of the case which includes the documents that belong to more than one category. The system used in the present study is a similar document search system that uses fuzzy clustering. The situation of belonging to more than one category for the documents is included by this system. The proposed approach consists of two stages to solve multicategories problem. The first stage is to find out the documents belonging to more than one category. The second stage is the determination of the categories to which these found documents belong to. For these two aims @a-threshold Fuzzy Similarity Classification Method (@a-FSCM) and Multiple Categories Vector Method (MCVM) are proposed as written order. Experimental results showed that proposed system can distinguish the documents that belong to more than one category efficiently. Regarding to the finding which documents belong to which classes, proposed system has better performance and success than the traditional approach.

Reverse modeling of a diesel engine performance by FCM and ANFIS

The paper includes reverse modeling of a diesel engine performance and emission characteristics. Modeling is done by fuzzy clustering method (FCM) and Adaptive Neural Fuzzy Inference System (ANFIS). Firstly, outputs and inputs parameters of a diesel engine were replaced as part of system. Later, these parameters were grouped into optimal numbers independently by using FCM and K-means clustering algorithm. Later on, these optimal numbers of clustered parameters were used as inputs and outputs of ANFIS to model engine performance and emissions characteristic. Input of the systems were power, torque, specific fuel consumption (sfc), nox, co2 and hc whereas outputs were air flow ratio, fuel rate, pboost, load and cycle. It has been seen that the best results obtained from ANFIS system by using FCM. What the proposed system makes different from pioneers are to be first study of reverse modeling and finding results as intervals instead of points. One more thing is that the load factor has never been implemented in previous studies but included in this study. Last but not least, the proposed system finds outputs in correct optimal interval as 100% ratio by FCM clustering and ANFIS.

Modeling the performance and emission characteristics of diesel engine and petrol-driven engine by ANN

In this study, performance and emission characteristics of an internal combustion (IC) diesel engine and petrol-driven engine were modeled by Artificial Neural Network (ANN). Diesel engine input parameters are air flow rate (Aflr), boost pressure (Pb), fuel rate (Frt), cycle (Cy) and load (L) whereas input parameters of the petrol-driven engine are advance (A) and cycle (Cy). Engine torque (Tq), power (P), specific fuel consumption (Sfc), emission values such as HC, CO2 and NOx of diesel engine and engine torque (Tq), power (P), specific fuel consumption (Sfc) and HC of petrol-driven have been investigated. R square values of Tq, P, Sfc, HC, CO2 and NOx of diesel engine were %99.9, %99.45, %99.32, %99.84, %99.71 and %99.26 respectively when ANN was used for modeling. R square values of Tq, P, Sfc and Hc of petrol-driven engine %97.24, %99.56, %98.19 and %97.19 respectively. The back-propagation learning algorithm with Hyperbolic tangent activation functions (for hidden layer neurons and output neuron) and 5:12:1 combination have been used in the topology of the network of diesel engine. The back-propagation learning algorithm with Logistic-Hyperbolic tangent activation functions (hidden layer neurons and output neuron) and 2:6:1 combination have been used in the topology of the network of petrol-driven engine. After having statistical t-test for outputs of both ANN, it has been seen that the obtained results are approximately %99.5 and %98.5 consisted (matched) with experimental data of diesel and petrol-driven engine. Main contribution of this work includes; 1) Dynamic load value was used as input parameters for diesel engine and so engine performance modeling and emission characteristic determination were done by regarding changing load, 2) The highest prediction values of output parameters are reached for both engine type regarding to the previous studies and 3) None of the previous studies include modeling of diesel and petrol-driven engine.

Analysis of ERASMUS staff and student mobility network within a big European project

The academic mobility is one of key factors that enable the globalization of research and education. In this paper we study the network of ERASMUS staff and student exchange agreements between academic institutions involved in FETCH - a big European project oriented towards future education and training in computer science. The structure of the network was investigated relying on standard metrics and techniques of social network analysis. Obtained results indicate that the network is in a mature phase of the development in which none of the institutions has a critical role to the overall connectedness of the network. Additionally, the network has a clear core-periphery structure with an active core and mostly inactive periphery.

Comparison of LSB image steganography technique in different color spaces

The incredible progress of technology has made the use of communication and information technologies indispensable because of the possibilities it offers. These possibilities increased the security issues on personal information and communication security problems such as phone calls, retrieving e-mail contents, copying private information on computers. Encryption algorithms used in classical security approaches, while ensuring the confidentiality of information, cannot provide the principle of “imprecision” that has become increasingly important in recent times. A coded or encrypted text can be solved by advanced machines when focused on it. So the key point is “do not raise suspicion”. For this reason, steganography and watermarking methods that put the invisibility of the existence of a secret message into the primary goal are especially the focus of interest after 2000s years. In this study, Least Significant Bit (LSB) technique, which is the most basic and commonly used technique in steganography, was applied to 3 different images in different color spaces. When the obtained results were compared according to the image quality evaluation criteria, it has been seen that the images in Hue-Saturation-Intensity (HSI) color spaces had better performances and successes to the other color spaces.

Performance and emission optimization of diesel engine by single and multi-objective genetic algorithms

In this study, single and also multi-objective (MO) genetic algorithms (GAs) were used for optimisation of performance and emissions of a diesel engine. Population space and initial population of both GAs were obtained by Artificial Neural Network (ANN). Specific fuel consumption (Sfc), NOx, power (P), torque (Tq) and air-flow rate (Afr) were reduced to %7.7, %8.51, %30, %4 and %7.4 respectively whereas HC increased at the rate of %10.5 by traditional single objective GA. HC, CO2, P and Sfc were reduced to %17.6, %30.05, %31.8 and %14.5 respectively whereas NOx increased at the rate of %13 by using multi-objective GA with Nondominated Sorting Genetic Algorithm II (NSGA II). %14.5 fuel reduction against %31 power reduction have never been obtained in the previous studies. This shows the effective usage of MOGA with NSGA II in optimisation of fuel diesel engine performance parameters.

Parallel inference in dynamic decision support systems of a semiotic type

The organization of parallel inference in dynamic decision support systems (DDSS) of a semiotic type, oriented towards a solving of ill-formed problems in dynamic applied domains, is considered. As a knowledge representation model, there are used production rules reflecting expert knowledge about a problem domain, an environment and decision making processes. The main concepts and assertions defining possibility and impossibility of parallel executing the production rules are given. Several types of parallelism in an inference process are introduced. The corresponding algorithm of parallel inference is described. Thus, the purpose of this paper is to develop and to research parallel inference methods and procedures that provide efficient processing a large amount of production rules for DDSS of a semiotic type.

An Investigation of Students, Parents and Teachers’ Satisfaction in Science High Schools in Konya Province in Terms of Some Variabilities

Education is the most decisive factor in the success of people in life and work. Today expectations in education have changed. Increases in education levels and facility of ways to access information have differentiated our level of social consciousness. Educational expectations of parents and teachers have also changed. There is now a mass who feign reluctance more and have a higher expectation from the school. Teachers’ expectations of students are also increasing. Researches examining the effects of school variables on student achievement have increased in recent years. Many studies indicate that schools with desirable characteristics have positive effects on student achievement. There are a large number of components that make the school environment come to life. Classroom sizes at schools, school culture, teaching methods used by teachers and professional qualifications are some of them. These components affect the satisfaction levels of teachers and parents working in schools. The aim of this research is to examine the satisfaction of teachers, students and students’ parents in Meram Science High School, Selcuklu Science High School and Karatay Science High School in terms of some variables in Konya province. Satisfaction scale was developed by researchers. Conducting a literature survey,the researchers have found that there were 46 questions and 3 open ended questions for the Student Satisfaction Scale, 45 questions in 3 sections including 5 demographic questions, 37 questionnaires and 3 open ended questions for the Parent Satisfaction Scale; For the Teacher Satisfaction Scale, 3 demographic questions, 43 scale questionnaires and 5 open-ended questions,the scales were finalized with 51 questions in 3 sections. When the answers given by the students are examined, the school satisfaction ratings were determined as undecided. When the answers given by the parents were examined, the school satisfaction level was partially determined as agreeing. When the answers given by the teachers were examined, the school satisfaction level was determined as strongly agreeing.

Comparision of numerical technique and artificiall intelligence techniques for performance modelling of a diesel engine

Comparision of numerical tehnique and Al techniques for determination of performance and emission characteristics of a diesel engine has been done in this study. Three different techniques namely multiple regression analysis, adaptive neuro-fuzzy inference system (ANFIS) and artificial neural network (ANN) were used for modeling aims. Engine torque (Tq), power (P), specific fuel consumption (Sfc), emission values such as HC, CO2 and NOx have been investigated. R2 values of Tq, P, Sfc, HC, CO2 and NOx were obtained as 99.9, 99.45, 99.32, 99.84, 99.71 and 99.26 respectively when ANN was used. Main contribution of this study includes; 1) First study that makes comperision between a numerical technique and Al tehniques. 2) Dynamic load value was used as input parameter. So that both engine performance modeling and emission characteristic determination were done regarding to changing load. 3) Highest prediction for values of output parameters were reached.