Nevcihan Duru

Improved Fuzzy Art Method for Initializing K-means

The K-means algorithm is quite sensitive to the cluster centers selected initially and can perform different clusterings depending on these initialization conditions. Within the scope of this study, a new method based on the Fuzzy ART algorithm which is called Improved Fuzzy ART (IFART) is used in the determination of initial cluster centers. By using IFART, better quality clusters are achieved than Fuzzy ART do and also IFART is as good as Fuzzy ART about capable of fast clustering and capability on large scaled data clustering. Consequently, it is observed that, with the proposed method, the clustering operation is completed in fewer steps, that it is performed in a more stable manner by fixing the initialization points and that it is completed with a smaller error margin compared with the conventional K-means.

Soil productivity analysis based on a fuzzy logic system

BACKGROUND Maintaining soil productivity is essential if agriculture production systems are to be sustainable. However, there is a paucity of tools for measurement for the purpose of understanding changes in soil productivity. Fuzzy logic-based analysis offers this possibility. It is a new method on the evaluation of soil productivity in Turkey and even in the world. RESULTS Values for pH, salinity, carbonate and organic matter were entered into the system as input variables so as to obtain soil productivity as the output. After the membership functions related to input and output were determined, rules were created. Then, the fuzzy logic system was applied separately to pH, salinity, lime and organic matter values of different soil types present in the Kocaeli region with the aim of obtaining corresponding fuzzy values. Thus, soil productivity profiles of the region were deciphered. CONCLUSION Organic matter levels in the study field remained below 30 g kg(-1) and varied between 22 and 28 g kg(-1). Productivity values were obtained as a percentage and varied between 16.9% and 18.1%. The lime content of the study soils varied in the range of 33-88 g kg(-1). Average totals for salt values of the field changed between 0.58 and 0.77 g kg(-1).

PREDICTION OF MAGNETIC SUSCEPTIBILITY CLASS OF SOIL USING DECISION TREES

Predviđanje magnetske osjetljivosti tla primjenom dijagrama za donosenje odluka Izvorni znanstveni clanak Magnetska osjetljivost (MS) je konstanta nedimenzijske proporcionalnosti koja pokazuje stupanj magnetizacije materijala u magnetskom polja. U nasem radu, cilj je predviđanje klasifikacije magnetske osjetljivosti tla primjenom algoritama za dubinsko istraživanje podataka (dobivanje korisnih, ranije nepoznatih podataka racunarskom analizom velikih baza podataka). Vrijednosti magnetske osjetljivosti ovise o sastavu, velicini zrna magneticnih minerala i njihovih izvora, litogenicnog, pedogenicnog i antropogenicnog porijekla. U radu smo primijenili dva d klasifikacijska algoritma za dubinsko istraživanje podataka nazvana ID3 i C4.5 za predviđanje vrste MS i stupnja zagađenja u podrucju Izmir u Turskoj. Primjenom algoritama, dobivaju se moguce vrste MS prema vrijednostima koncentracije teskih metala (Pb, Cu, Zn, Co, Cd, Ni). Cilj primjene algoritama je izrada dijagrama i pravila za donosenje odluka u svrhu dobivanja vrijednosti MS. Na taj nacin, eliminiraju se greske nastale promjenom uvjeta okoline i teskoca u mjerenju. Prema tim pravilima, dobili smo uvjete tocnosti od 82 % i pokazali da su vrijednosti ispitivanja i vrijednosti mjerenja međusobno kompatibilne. Kljucne rijeci: dubinsko istraživanje podataka; klasifikacija; magnetska osjetljivost; zagađenost teskih metala

A novel method for software effort estimation: Estimating with boundaries

Software effort estimation is a crucial phase in software project management. Accuracy of estimation directly affects project success or failure. Managers try to estimate proper effort resources and this is a challenging issue for management. Having a set of tools and methodologies, estimation process can be made better. COCOMO is one of the most used model which has a parametric form. Also, artificial neural networks (ANN) are combined with COCOMO and these methods increased overall performance. However, effort estimation process generally produces one output; estimation value. It is a well-known issue that a project manager must keep in the mind that any estimation must have some upper and lower limits, boundaries. In this paper, a novel method, combining COCOMO used ANN with K-Means is used to estimate effort and possible boundaries. ANN output is used as input to K-Means sets and proper set value is calculated, including possible lower and upper effort estimation value. Experimental results are shown that proposed method has acceptable results over ANN and COCOMO.

The New Approach of AES Key Schedule for Lightweight Block Ciphers

This paper considers block ciphers and key schedule algorithm that is one of the crucial components of a block cipher. It computes round keys/subkeys for relevant round from a short key. The presented experiments show that proposed key schedule algorithm which inspired by Advanced Encryption Standards (AES) key schedule has desirable properties: Avalanche Effect and Strict Avalanche Criterion (SAC). It satisfies good bit confusion and diffusion. The average success rate of the proposed key schedule algorithm for the SAC test is 95%. As a side result it was found that while testing SAC effect computed values that lie between confidence lower and upper bounds, greater than upper bounds and less than lower bound all of them reach normal distribution. Also based on example given experimental result, proposed structure exhibits a very strong Avalanche Effect because almost at the first round approximately half the bits are changed in the key.

Automatic lung segmentation by using histogram based k-means algorithm

In this study, it was developed an histogram based k-means algorithm by considering the region features in order to segment whole lung region automatically. After the segmentation the lung lobes are extracted from the surrounding tissue. This method could be used as a first step of various computer aided analysis of lungs. In the study, 34 cases which were scanned by the three different tomography systems, the processed. The method provides 96% segmentation success rate without performing normalization process. The histogram values have been used to calculate the distance between the points instead of using the cartesian system of the traditional k-means method. Thus, similar tissues which are not connected and far from each other, were included in the same set. The histogram based k-means algorithm was compared with Fuzzy C-means (FCM) and optimal thresholding method, found more efficient for the iteration number and segmentation accuracy.

An improved method for fuzzy clustering

Adaptive Resonance Theory (ART) is an unsupervised neural network. Fuzzy ART is a variation of ART, allows both binary and analogue input patterns. However, Fuzzy ART has the cluster overlapping problem. In this study, to solve this problem, we propose a new Improved Fuzzy ART (IFART) algorithm. In the proposed algorithm, after the clusters are formed, membership degrees of each data instance to all clusters are calculated according to the cluster centers. If data instances are not in the cluster with maximum membership degree, then they are moved between clusters according to their maximum membership degrees. The clustering results on real sample datasets are investigated and compared with the conventional Fuzzy ART. It is seen that, Improved Fuzzy ART is more efficient then Fuzzy ART and also a high performance algorithm than SOM.

Prototype of a tool for analysing laryngeal cancer operations

In this study, a software tool was developed to analyse the medical data collected from laryngeal cancer operations by using two data mining techniques. The software, run on real-world medical data, is a tool that enables medical decisions to be reached by analysing past records from patients. The k-means algorithm, which is a clustering algorithm in data mining, was used to point out the intensities in the data set and to display two dimensions on the charts. The data of three screens that were named as selective clustering, different pre- and post-operation stages and clustering operations based on pre-operation T values, were processed using clustering with the k-means algorithm and one screen, which named relapse and survival percentages, was processed through classifying. It helps the future decision-making process by considering false estimates of pre-operation stages of the cases and by using the information gathered from past cases concerning tumour relapse and the survival percentage for prognostication. The characteristics of laryngeal cancer operations data, that involve causal links, were exposed by using two data mining techniques in this application.

A novel approach about cohesion measurement for classes

Cohesion refers to the degree of the relationships among the members in a class. A class is cohesive when its members are highly correlated. Several metrics have been proposed in the literature in order to capture class cohesion in terms of connections among members. They generally count the number of attributes used by methods or the number of methods pairs that share attributes. They constitute a restrictive way for capturing the cohesion. Because they do not consider some characteristics of classes like that special methods, disjoint interaction patterns and connectivity among class members. In this study, a new criterion, which focuses on interactions and groups between class members with considering density of connections among members and incorporates the special methods to cohesion capturing process, is presented, and a new notion about determination of class cohesion is proposed.

Fuzzy logic based intelligent tool for databases

In this paper, a software tool for enabling fuzzy query from a classical database is introduced. By using this tool, some fields (attribute) of a database table can be fuzzified and a supplementary database, which includes fuzzy values, is formed. Developed software tool is applied in a sample database including some fields about the students for the evaluation of scholarship application. It is concluded that, the fuzzy query method is more flexible and the results of such query are more predictive.