Search

In this paper, we present a web-based optical character recognition (OCR) system that converts images of Ottoman documents printed with naskh font into text using CNN+RNN-based deep neural network models. For training, three datasets - original, synthetic, and hybrid - were prepared and three different OCR models were created. The original data set consists of 1,000 pages and the synthetic data set consists of 23,000 pages. Hybrid data set contains both. The trained models were compared with Tesseract's Arabic and Persian, Google Docs' Arabic, Abby FineReader's Arabic, and Miletos OCR model/tools with a 21-page test set. The comparison was made with 3 different texts (raw, normalized, and joined) and using 3 different criteria (character, ligature, and word recognition). The Osmanlica.com Hybrid model produced significantly better results than the others with 88.86% raw, 96.12% normalized, and 97.37% joined accuracy in character recognition; 80.48% raw, 91.60% normalized, and 97.37% joined accuracy in ligature recognition; and 44.08% raw and 66.45% normalized accuracy in word recognition. To investigate the effects of the characteristics of the alphabet on OCR, character, ligature, and word frequency analyses of Ottoman was performed. In this analysis, the characters in the alphabet were grouped according to distinctive features such as connectedness, letter body, position and number of dots, type of character, and source language; and frequencies and recognition accuracies were examined for each group. OCR results are also reported for each character. [ABSTRACT FROM AUTHOR]

Intelligent transportation systems use parameters such as traffic flow, density and speed to manage city traffic. This paper presents a novel prediction model for traffic speed prediction consisting of nine stages. In the presented model, real vehicle data were passed through data filtering and map matching processes, density-based clusters were created, cluster features were generated, instant traffic state was displayed, and traffic speed prediction was performed using with the artificial neural network RNN model. In previous studies, while traffic speed prediction can be performed on a specific road with stationary data sources or on different days with distributed GPS records, in the developed model, characteristics of the interested road are created by obtaining density-based vehicle cluster features, short-term and data-driven speed prediction is made within the changeable structure of traffic. Speed prediction was tested on Eskisehir and Istanbul roads belonging to Ankara province, the error rates were determined for speed prediction using the RNN variant LSTM and GRU methods, Eskisehir road LSTM-GRU error rates were measured as 8,595-8,656 and İstanbul road error rates as 7,331-7,955, respectively. The developed model for the changeable nature of traffic has yielded successful results in near real time. It is considered that the proposed model will offer different and new solutions in the prediction of traffic parameters, accelerate the processes and assist to the users more accurate and faster services. [ABSTRACT FROM AUTHOR]

In this paper we classified 4 skin lesions (Melanoma, Melanocytic Nevus, Basal Cell Carcinoma, Benign keratosis) from ISIC 2019 dataset which was published by International Skin Imaging Collabration in 2019. We used InceptionV3 convolutional neural network model for classification. We applied two preprocessing methods: High Dimensional Model Representation (HDMR) and Hilbert Transform. In conclusion we obtained 89% accuracy on classification of Basal Cell Carcinoma using Hilbert Transform. Moreover, we obtained 78% accuracy on classification of Melanoma using Contrast Enhancement High Dimensional Model Representation (HDMR). [ABSTRACT FROM AUTHOR]

Social media users have been growing exponentially in recent years. This growth has evoked researchers and manager to analyze the social media and customer sentiment because it offers significant opportunities to advance business intelligence in supply chain. However, supply chain members are struggling in understanding the general sentiments in today’s business world. Therefore, various methods are used to provide valuable insights in supply chain. In this paper, SentiStrength is used to analyze customer reviews related to one type of product. The output of SentiStrength and demands of the product are then fed into Artificial Neural Network to forecast the customer demands. After, nondominated sorting genetic algorithm II (NSGA-II) based simulation optimization is employed to solve the inventory routing problem using forecasted customer demands. The results of the study demonstrated that the use of hybrid methodology containing sentiment analysis can successfully analyze the inventory routing problem. [ABSTRACT FROM AUTHOR]

In this paper, a new hyperspectral image classification method based on 2-dimensional complex Gabor filtering and deep convolutional neural networks is proposed. Specifically, as a deep learning model, convolutional neural network is aimed to extract distinctive high-level features. Deep-learned and Gabor feature extraction methodologies are simultaneously performed on the input hyperspectral samples. Gabor features are calculated by implementing complex Gabor filtering only on the first three principal components of the hyperspectral image. The proposed hybrid model uses Gabor transform to obtain local image features, such as edges, corners and texture. The Gabor features of the images are calculated at multiple orientations and frequencies. Then, deep features and Gabor features are fused to obtain a more robust and discriminative feature vector. Hybrid feature vector is used as input to a softmax classifier for hyperspectral image classification. The parameters of the proposed deep learning architecture are optimized using a small training set. Thus, the over-fitting problem of the proposed convolutional neural network has been reduced to some extent. Experiments performed on two popular hyperspectral datasets show that the proposed method can achieve better classification performance than some conventional methods. Classification results demonstrates that the proposed hybrid model is an efficient method for feature extraction and classification of hyperspectral images. [ABSTRACT FROM AUTHOR]

Leveraging machine learning techniques in NLP domain has been a very hot research field due to the advancements in artificial intelligence area. Despite the popularity of this field, there is no known study on application of ML techniques on old Turkish language. This study aims to fill in this gap where 32000 pages of text has been downloaded from the websites of Ministry of Culture and a two-layer neural network model has been built on top of them to discover the semantic similarities between Turkish words in old Turkish language. The algorithm has been run with different parameters such as window size, dimension size, sampling size etc. and the produced vector spaces are uploaded into public servers for the purposes of enabling a RESTful API based query interface. Also a web UI has been created to provide a querying mechanism for regular users. The services that are developed can be used for two different purposes. One of them is to integrate these services into existing old Turkish language dictionary websites that are made available by third party providers as well as other institutions such as Ministry of Culture and Turkish Language Institution. Secondly, the developed services are intended to be used for mitigating the translation errors made during the translation of old Turkish texts into modern Turkish language in the areas of history and Turkish literature. Also enabling these services for public use will encourage other researchers to pursue this academic work and compare their results with the experimental results presented in this paper to make further improvements in this field. [ABSTRACT FROM AUTHOR]

In this study, determinants of second-hand automobile prices are examined and focused on the estimation of price using the second-hand automobile advertisements in web sites. As known, second-hand automobile pricing is a difficult matter for both automobile buyers and sellers. The primary goal of this paper is to identify the determinants of second-hand automobile pricing. Additionally, the second purpose is to compare the prediction performances of two statistical models in order to predict second-hand automobile prices. In this context, hedonic models and artificial neural networks are handled in this study. Hedonic models utilize multiple regression models on large data sets in the analyses. On the other hand, artificial neural networks are employed in this studty as an alternative method on account of potential non-linearity in the hedonic functions. As a conclusion, the present paper demonstrates both determinants of second-hand automobile prices in Turkey, and also artificial neural networks can be a better alternative method for prediction of the second-hand automobile prices. [ABSTRACT FROM AUTHOR]

This paper aims to present technical information about the heat transfer mechanisms in human body, reaction of the human body temperature to the environment (thermoregulation), measurement of the body core temperature, and the action to be taken as the body temperature changes abnormally. Detailed information about the diseases due to abnormal temperature changes in the human body (i.e., hypothermia, hyperthermia) and about methods to prevent these diseases are reviewed. This paper is a presentation of fundamental information on bioheat transfer and thermoregulation for the interested researchers, and it is hoped to be a guide for engineering researches which aim to simulate human body. [ABSTRACT FROM AUTHOR]

Main focus of this work is to realize image searching within another image in an efficient way. Image searching within another image is accomplished through the integrated use of image matching techniques and searching algorithms. Artificial neural networks along with various image features such as average color value, color standard deviation, correlation and edge parameters are used for image matching whereas genetic algorithms were used for image searching. In the work presented in this paper, an integrated method based on smart searching algorithms, quick image matching methods and parallel programming techniques were proposed and implemented. Proposed method was tested on several low and highresolution reference and template images. Results revealed that the proposed method can successfully match images and significantly reduce the total search time. [ABSTRACT FROM AUTHOR]

Highly accurate prediction of high performance concrete (HPC) compressive strength is very important issue. In recent years, a variety of modeling approaches and methodologies have been applied to predict HPC's compressive strength from a wide range of variables, with different ratios of success. In this study, an appropriate machine learning method, using different mixing ratios for the prediction of compressive strength of HPC, is investigated. In recent years, rather developing machine learning methods; Artificial Neural Networks (ANN) and Support Vector Machines (SVM) 's applicabilities for the prediction, handled in this study, are being investigated and extremely high results were obtained. In this paper, it's obtained that prediction success of SVM has been found more satisfactory than ANN's. It is concluded that the SVM's can be used effectively as an alternative method by research labs and the concrete firms for predicting the strength. [ABSTRACT FROM AUTHOR]

Fraud is an intentional action in the financial statements. Fraud can be defined as an act of deception where an individual or a group obtains benefits in return for damaging another individual or group. Frauds can be classified as employee fraud and the fraudulent financial reporting. According to the 2012 report of Association of Certified Fraud Examiners, companies lose averagely more than 1 million dollars due to financial statement fraud. Therefore, forensic accounting has an important role in determining financial statemet frauds in order to provide legal support in lawsuits. The purpose of this paper is show to use artificial neural network model detecting frauds in forensic accounting. For this purpose, we conduct an empirical research in the Borsa Istanbul. In the study, the correct classification of the artificial neural network is realized as 100%. [ABSTRACT FROM AUTHOR]

Forecasting studies are extremely important in the technical, social and economic research. Generally, we know it is very difficult to forecast with higher accurate about a system by using recent values. In the scientific literature, the forecasting studies of energy, personnel planning, production planning, climate changes, sales and marketing and economics etc. are frequently found. In this paper, for an emergency calls center in Isparta province of Turkey an artificial neural network (ANN) forecasting model was developed to determine the number of calls for as health, fire and security services on a pilot implementation of the emergency calls center on a single number 112. In the developed model, the gradient descent with adaptive learning and momentum (GDX) algorithm is selected as the training algorithm with feed-forward back-propagation by using 80% of input data and the 20% of input data is used for testing set data from last month. After the testing, the mean absolute percentage error (MAPE) rate is obtained as 4.5% and it is useful to test. In addition, the forecasting results of the next month are shown that the MAPE values are 2.65%, 6.40% and 5.24% with ANN, trend analysis and ARIMA (1 1 1) models respectively and, the number of calls are found separately on the types of calls in daily. Consequently, the developed model by using ANN to forecast the number of calls in an emergency call center is more accurate than the trend analysis and ARIMA models. [ABSTRACT FROM AUTHOR]

Forecasting studies are extremely important in the technical, social and economic research. Generally, we know it is very difficult to forecast with higher accurate about a system by using recent values. In the scientific literature, the forecasting studies of energy, personnel planning, production planning, climate changes, sales and marketing and economics etc. are frequently found. In this paper, for an emergency calls center in Isparta province of Turkey an artificial neural network (ANN) forecasting model was developed to determine the number of calls for as health, fire and security services on a pilot implementation of the emergency calls center on a single number 112. In the developed model, the gradient descent with adaptive learning and momentum (GDX) algorithm is selected as the training algorithm with feed-forward back-propagation by using 80% of input data and the 20% of input data is used for testing set data from last month. After the testing, the mean absolute percentage error (MAPE) rate is obtained as 4.5% and it is useful to test. In addition, the forecasting results of the next month are shown that the MAPE values are 2.65%, 6.40% and 5.24% with ANN, trend analysis and ARIMA (1 1 1) models respectively and, the number of calls are found separately on the types of calls in daily. Consequently, the developed model by using ANN to forecast the number of calls in an emergency call center is more accurate than the trend analysis and ARIMA models. [ABSTRACT FROM AUTHOR]

In mass customization using resources efficiently with changing customer demands and manufacturing conditions is vital, since in today's competitive market conditions, this efficiency provides cost, time and labor savings to companies. In production environment, Process Planning (PP) and Scheduling are two functions that provide efficient usage of resources. Isolation and long time gap between these functions are main problems affecting the effiency of production. In this study, to solve the problems of productivity, it is aimed to build an integrated system which is able do PP and Scheduling in parallel and respond quickly to fluctuations in job floor. In existing integration models aiming to eliminate this isolation and time gap, it has been observed that if the search and solution space expands, the computational time increases rapidly. Therefore, in this research, a hybrid optimization approach, which can find the optimal solution rapidly is considered and a hybrid model combining both Genetic Algorithm (GA) and Artificial Neural Network (ANN) is proposed. To improve GA performance and increase the effiency of searching, clustering activities are carried out for building new cromosome structures. To increase population diversity, effective genetic operator schemes and efficient genetic represantations are used. In the integration module, 3 different GA structures created within the scope of our research are compared and the algorithm formed by clustering method shows better performance than the others. In this paper by using ANN method, a new system trained by data obtained from Scheduling is generated and this system is able to quickly respond the changes in shop floor and provide new schedules instantly. In rescheduling module, ANN's performance measures provide evidence to accuracy of Heuristic solution generated by Integration module. [ABSTRACT FROM AUTHOR]

The use of artificial neural networks (ANN) in geotechnical engineering has gained wide application in Turkey as well as the world during the past ten years. A comprehensive literature survey has shown that applications are concentrated in basic areas such as classification, site characterization, liquefaction, hydraulic conductivity, compaction, consolidation as well as practice comprising the problems of retaining structures, settlement of foundations, pile capacity and modeling of soil behavior where the relation-ship among the several parameters involved is not always thoroughly understood. Several case histories are presented as examples. Latest research shows that artificial neural networks are heading towards unison with fuzzy logic and genetic algorithms and it is certainly superior to the statistical methods . The second part of the paper gives an account of research conducted into estimation of the soil profiles in the city of Adapazari, Turkey. There has been widespread damage and destruction in the city during the Mw=7.4 earthquake in 1999. The damage has largely been blamed on inferior alluvial deposits and parts of the city have been moved to the North where the soil was found to be "sound". The sediments in the city are the products of the meandering river Sakarya which also flooded the region almost biannually. The two processes have resulted in the formation of complex soil profiles and near chaotic profiles are frequent. The silty layers are possibly the source of ground failures, occasionally leading to liquefaction in the absence of sands. A comprehensive soil investigation has been carried out since 1990 by boreholes and cone penetration soundings. The authors have used the rich database available established from previous and current laboratory and field investigations. Out of this voluminous data those pertaining to depths of 2 to 7m have been used for the ANN work, as those depths have been diagnosed as the possible liquefaction zone. Data from 117 CPT sites whose coordinates were known were employed for this study. The 3236 readings of tip resistance and sleeve friction were used to establish the ANN model. The well established Robertson classification chart defines nine types of soil. It requires the normalised values of tip resistance(Qt) and and sleeve friction(Fr) a identify the soil layer. The Qt* and Fr* are further defined to form the spatial distribution by the use of equations. The training matrix even for a limited depth of 5m using the data from 90 CPTU tests came out to be of size 3236x3236, which was difficult to handle. Consequently, thirteen 1800 by 1800 matrices were established (1.60-1.98m, 2.00-2.38m, 2.40-2.78m, 2.80-3.18m, 3.20-3.58m, 3.60-3.98m, 4.00-4.38m, 4.40-4.78m, 4.80-5.18m, 5.20-5.58m, 6.00-6.38m, 6.40-6.78m, 6.80-6.98m). Data from 27 CPTU were used to form the thirteen 540 by 1800 simulation matrices and 1800 by 1800 training matrices. Analyses were carried out on the Matlab 2010a Toolbox7 NNtraintool interface. 60% of the data were employed for Training, 15% for Validation and 15% for Testing. Inspecting the results, it was found that the success rate in estimating the soil profile anywhere in the 26 km² city area was as high as 92%. This is a surprisingly high success rate considering the highly complex and laterally variable soil profiles throughout the city. [ABSTRACT FROM AUTHOR]

Because of increasing need for reliability, safety, maintainability and survivability in technological systems, fault tolerant control system has become extremely important. Fault tolerant control systems have been developed to overcome some weaknesses of the conventional feedback control design, such as instability and unsatisfactory performance in the faulty cases. In complex systems such as aircrafts, nuclear power plants, chemical plants etc., the results of a minor fault in the system can be destructive. Therefore, it is necessary to design control systems that are able to tolerate potential faults in these systems. A control system with this kind of fault tolerance capability is defined as fault tolerant control system. A fault tolerant control system is a closed-loop control system which can tolerate malfunctions while maintaining desirable performance and stability properties. A typical active fault tolerant control system includes three subsystems: reconfigurable controller, fault detection and diagnosis unit and decision supervisor unit. The reconfigurable controller is placed in control loop, instead of regular controller in traditional closed-loop control system. Since the real time toleration of faults can only be achieved if the control system has an automatic reconfiguration mechanism. This reconfiguration is generally activated after a fault has been detected and isolated. A fault detection and diagnosis system is a unit that obtains the occurrence of faults and determines their features in terms of type, location, size and/or time. The decision supervisor activates reconfiguration action in response, which can be pre-determined for each fault or obtained from real time analysis and optimisation. In active fault tolerant control systems, reconfiguration mechanism can be classified as on-line controller selection and on-line controller calculation techniques. In the on-line controller selection approach, the controllers associated with presumed faulty conditions are computed in an off-line manner in the design stage and they are selected in an on-line manner based on the real time information from fault detection and diagnosis algorithm. In the on-line controller calculation approach, the controller parameters are calculated in an on-line manner after the occurrence of fault. In this study, an active fault tolerant control technique, support vector machines based direct fault tolerant control system, is presented. In general, reconfiguration mechanism of a fault tolerant control system utilizes the information from fault detection and diagnosis unit at the decision stage. In the presented method, reconfiguration mechanism and diagnosis unit work independently. Both of them use only real time system outputs. A powerful and fast learning algorithm is needed for this purpose. Support vector machines is one of the most popular intelligent machine learning tools. They have become a very good alternative of neural networks with their superior generalization capacity, classification, regression and modeling performance. In this paper, support vector regression machines have been used in fault detection and diagnosis process and also in reconfigurable controller unit. PID type controllers have been used in reconfiguration sub-system. The PID coefficients of faulty and un-faulty cases to be used in training stage are obtained by genetic algorithm approach in an off-line manner. In reconfiguration mechanism, for each PID coefficient one support vector regression machine is set up. Faulty and un-faulty system outputs are collected for different input signals. In training stage, faulty system outputs and corresponding PID parameters are used as support vector regression machines' inputs and outputs, respectively. Thus the training phase of support vector machines is completed in an off-line manner. After completing the training phase, system is run for the reference input signal. The outputs of system are sent to decision unit periodically. Three of support vector regression machines are simultaneously evaluated the data sent by the system, and produce coefficients of the PID controller. The controller of which its coefficients are reconfigured, starts to work to maintain system performance in an on-line manner. In order to determine the type of fault, a similar process is exploited using one support vector regression machine. In training phase, inputs of this support vector regression machine are faulty and un-faulty system outputs but outputs are the features of the faults. This fault diagnosis unit runs parallel with intelligent fault tolerant controller. The performance of this knowledge based fault diagnosis and active fault tolerant control methods is illustrated on simulation example involving a two-tank water level control system under faulty conditions. [ABSTRACT FROM AUTHOR]

In this study, Turkey's Carbonization Index (v) depending on the basic energy indicator has determined as an analytical equation. In this paper, artificial neural networks (ANN) which is quite often in recent years and the possibility of applying to many fields, are used to obtain analytical expressions. Carbonization Index is defined the amount of CO2 emissions per energy consumption. This parameter is an important for Turkey because of has signed the Kyoto protocol, February 2009. According to the results, accuracy of the carbonization index formulated empirical was obtained as R²= 1. It is expected that this study will be helpful in demonstrating carbonization index for policy makers. [ABSTRACT FROM AUTHOR]

Electrocardiogram is widely used in cardiology since it consists of effective, simple, noninvasive, low-cost procedures for the diagnosis of cardiovascular. The state of cardiac heart is generally reflected in the shape of ECG waveform and heart rate. Cardiac arrhythmia is any of a group of conditions in which the electrical activity of the heart is irregular or is faster or slower than normal. There are many inherited conditions and heart diseases that can cause ECG signals arrhythmia occurrence. The one of the most difficult problem faced by to-day's automatic ECG analysis is the large variation in the morphologies of ECG waveforms. The ECG waveforms may differ for the same patient to such extend that they are unlike to each other and at the same time alike for different types of beats. In last decades, cluster analysis, has been combined with other techniques, has been used to overcome these difficulties in many areas of ECG processing, such as classification of ECG arrhythmias, ECG feature selection, ECG character points detection, classification of ECG morphology. In this paper, Ant Colony Optimization (ACO) based clustering analysis of ECG arrhythmias taken from the MIT-BIH Arrhythmia Database is proposed. Both time domain and discrete wavelet transform based frequency domain features are used in the analysis. Mainly, three operations must be implemented for computer aided classification: 1) preprocessing 2) feature extraction, 3) classification. Preprocessing stage contains filtering, baseline detection and correction, the detection of ECG signal's characteristic points and normalization. Feature extraction is the process in order to deter-mine the different coefficients for describing the ECG waveform. In this work, in order to extract the best features that will represent the structure of ECG signals, methods based on time domain features and Discrete Wavelet transforms are used. Since the numbers of wavelet coefficients are huge amount as compared to the time domain parameters, principle component analysis based compression is applied on them in order to decrease their number to the same level of time domain features. Then, the reduced number of frequency parameters is combined with the time domain features in order to get the total feature sets. Different types of feature sets are tried and the classification results are com-pared. These are; rare time domain feature set, rare frequency domain feature set and the mixture of them. Third stage is the classification of ECG signals according to their feature sets. In this thesis, Ant Colony Optimization (ACO) based clustering analysis of 6 different type arrhythmias taken from the MIT-BIH Arrhythmia Database is proposed. These are normal sinus rhythm, premature ventricular contraction, atrial premature contraction, right bundle branch block, ventricular fusion and fusion. Both time do-main and wavelet based frequency domain features are used in the analysis. In ACO method, the k-nearest neighborhood classification algorithm is used to determine the ECG signals' class. Kohonen and back propagation neural network algorithms are developed in parallel to verify and measure the ACO classifier's success. The ACO based classification method is depends on a learning method that can be called semi-supervised. In the stage of construction of intra-class cluster, the label of input sample's class is known but the clusters are determined by algorithm. Therefore the proposed system's success is com-pared with the supervised back propagation neural network and unsupervised Kohonen networks. Be-cause the back propagation neural network is one of the most used methods in literature the comparison with ACO method is important also. Kohonen networks are is similar learning algorithm with ACO clustering. They are unsupervised net-works having competitive learning scheme with Euclidean distance based "winner-takes-all" rule. Classes of feature vectors in the input space are not known at the beginning. Nodes are determined then labeled by a labeling process. Comparisons results indicate that the mixture feature set give a better success for classification. [ABSTRACT FROM AUTHOR]

In the last decade, with the growth of computational facilities, there are many academic attempts about financial market prediction. Because of the huge sort of local and global data that may affect market trends, forecasting models developed until today have been only effective in few markets and for a limited period of time. Studies about Istanbul Stock Exchange (ISE) are limited. Determining the complex relationship between financial markets and economic parameters needs a chaotic approach. Chaos is a nonlinear deterministic process which only appears random because it cannot be easily expressed. Multivariate analytical techniques using both quantitative and qualitative variables have repeatedly been used to help form the basis of investor stock price expectations and hence, influence investment decision making. However, the performance of multivariate analytical techniques is often less than conclusive and needs to be improved to more accurately forecast stock price performance. A neural network (NN) method has demonstrated its capability of addressing complex problems. A neural network method may be able to enhance an investor's forecasting ability. There are several motivations for trying to predict stock market prices. But initial background needs of almost everyone is to investigate the market behavior, because each study about financial markets starts on that base. The most principal motivation is naturally financial gain. Finding any system that can consistently give greater revenue then the average return in the dynamic market place, is very important for the owner of the system, to give market players the impression of being trustworthy. Moreover, many investors are continually looking for this superior system which will yield them higher returns. Another motivation for researchers is the Efficient Market Hypothesis (EMH). It has been proposed in the EMH that markets are so efficient and profit opportunities are discovered so quickly that no one can catch superior performance. Hypothesis states that no system can continually beat the market because if this system becomes public, everyone will use it, thus negating its potential gain. There has been an ongoing debate on the validity of the EMH, some researchers use forecast results to validate their claims, other observers trying to demonstrate that the market is weak in efficiency. Validity of EMH is very important for the board managers point of view. If it can be proved that the market behavior is rational, while researchers evaluate the concept of a new efficient market, it would be possible to attract new resources to the market and create a new transaction system in which investors believe. No matter what the prediction purpose and methodology are the first step is to choose the parameters that really represent the market direction of movement and the level of variation. On the other hand, typical investors trading behaviors have been guided by two main perspectives; namely; Fundamental and technical analysis. While technical analysis uses only past market prices and volume, fundamental analysis is derived from external information which comes from the economic systemen circling the market. This information consists of interest rates, prices and returns of other assets, and many other macro or micro economic variables. The aim of this study was to evaluate the effectiveness of using macroeconomic indicators such as; gross national product, industrial production level and financial data such as; real benchmark, currency exchange rates, in predicting movements in the ISE 100 Index (ISE-100). Availability of midterm economic and financial indicators to estimate the ISE-100 trends was investigated in this research. If the investigation findings show that these variables are efficient to explain the price movements in the ISE, then two basic judgments will be reached. Firstly, the EMH will be canceled or at least in a most optimistic view, a weak form efficiency in the market will be proven. Secondly, an investment decision support system based on NN estimates will be developed for investment advisors and investors. Also, the EMH is presented and contrasted with chaos theory and neural networks. This paper refutes the EMH based on previous neural network work. Finally, future directions for applying neural networks to the financial markets are discussed. Using statistical investigation we find causal relations between some indicators including macroeconomic, inflation, alternative financial instruments, and the Turkish Security Market. After adjusting seasonal and influationist impacts; input variable has been investigated their midterm relationship with ISE-100 Index. In the last section, appropriate normalization method for a backpropagation feedforward networks (BPNN) was researched. [ABSTRACT FROM AUTHOR]

In this paper, the aim is to classify natural or crushed aggregates by using concrete and asphalt mixes through Artificial Neural Networks. For classification, it was a used the feature vector which was calculated by using digital image processing techniques. Of the five different type coarse aggregates images were taken with 45° and 90° by a 10 Mp (Sony DSC-R1) and 7.1 Mp (Canon EOS 350D) camera. Aggregates images were processed and analyzed by using MATLAB Image Processing and Neural Network Toolbox. Classification process was made with totally 18 feature vectors, which is 9 vectors each angles, by neural network. Results showed image processing and neural networks which are important methods for founding shape parameters and classification of aggregates, and performance, cost and time consuming factors of automation systems in aggregate sources will be effective with these methods. [ABSTRACT FROM AUTHOR]

In this paper, the aim is to classify natural or crushed aggregates by using concrete and asphalt mixes through Artificial Neural Networks. For classification, it was a used the feature vector which was calculated by using digital image processing techniques. Of the five different type coarse aggregates images were taken with 45° and 90° by a 10 Mp (Sony DSC-R1) and 7.1 Mp (Canon EOS 350D) camera. Aggregates images were processed and analyzed by using MATLAB Image Processing and Neural Network Toolbox. Classification process was made with totally 18 feature vectors, which is 9 vectors each angles, by neural network. Results showed image processing and neural networks which are important methods for founding shape parameters and classification of aggregates, and performance, cost and time consuming factors of automation systems in aggregate sources will be effective with these methods. [ABSTRACT FROM AUTHOR]

The purpose of this paper is to present an artificial neural network (ANN) based method for the estimation of company success during economic crisis periods. There are a number of studies about recovering the crisis successfully by companies. Success or failure interpretation can be made upon financial indicators of companies for the period before the crisis. These indicators also provide estimation about the success of the companies. In this study, an ANN model is proposed estimating the success of companies after crisis period by means of financial indicators of the companies for the year before 2001 economic crisis of Turkey. A prototype model is designed with regards to homogeneous data set retrieved from the companies facilitating in manufacturing sector. [ABSTRACT FROM AUTHOR]

In the everyday life of humans, to define the rules used to recognize a certain sound, image or an analog data necessitates a sequence of complex processes which sometimes becomes impossible to accomplish. However, to develop well defined software and hardware based criteria in the application of pattern recognition problems, is possible. The aim of classification can be defined as to assign an unknown object to a class containing similar objects (or to distinguish objects having the same properties from those not possessing). Classification is especially important in the real world applications or in the digital world. Basic classification methods using nearest neighbourhood algorithm have first been seen in early sixties under the subject tile" pattern recognition." Classification is used in a huge variety of applications such as automatic target identification, artificial neural networks, artificial intelligence, template matching, pattern recognition, analog to digital converters, quantization, medical diagnosis, statistics etc. Therefore nowadays, be it in the real or digital world, data classification is becoming increasingly important. But until recently, major work on classification was on developing algorithms used in software packages whereas, in many applications it is becoming more and more important to classify data much faster and in real time, entailing the need for hardware realization of these algorithms. Software approaches are not practical for real time applications, the processing is computationally very expensive, consuming a lot of Central Processing Unit (CPU) time when implemented as software running on general purpose computers. So in literature hardware implementation of classifier topologies become necessary. Also in literature hardware realized classifiers which are designed to work in low power operation; moreover some of these hardware classifiers do not have custom tunability. So they can only be used for a specific application. The recent developments in electronics technology has created a perfect medium for the hardware realization of classifier structures which, in turn, will render many classifier application prospects feasible in real time. This paper targets the design and application to real world problems of tunable, low power new classifiercircuits using CMOS technology. So, a low-power CMOS implementation of a multi-input data classifier with several output levels is presented. The proposed circuit operates in current-mode and can classify several types of analog vector data. An architecture is developed comprising a threshold circuit based on CMOS transistors operating in subthreshold region. To this purpose a one dimensional classifier, called core circuit is proposed. The core circuit also works as a one-dimensional classifier. As this circuit is designed to operate in currentmode the input and the output data is provided to the core circuit with currents.… [ABSTRACT FROM AUTHOR]

In this paper, within the framework of the Act on the Right of Information Acquirement, an electronic application classification system has been developed in order to eliminate such problems as the waste of workforce and time during the classification of electronic applications to the institutions. The system first classifies the electronic applications to Erzincan University then forwards them to the relevant Chambers of Affairs. Artificial Neural Network based system, which considerably accelerates the flux of work and hinders the waste of workforce, is a word based and supervised learning system and it works with 98% accuracy. [ABSTRACT FROM AUTHOR]

Artificial neural networks (ANNs) have become modeling tools that have found extensive acceptance and they have frequently used in applications in many disciplines for solving complex problems. Different ANN structures are valuable models, which are used in the medical field for the development of decision support systems. In this paper, the learning and classification processes are used for determining the level of bone-density (safe / risk of osteoporosis) in woman. In this study, three different structured neural networks were used for classifying of osteoporosis and the most efficient structure was determined. The training network structures were Multilayer perceptron neural network (MLP), Linear Vector Quantization (LVQ) and Self Organizing Map (SOM). Performance indicators and statistical measures were used for evaluating the structures and the results demonstrated that the MLP was the most efficient structure for classifying of osteoporosis. [ABSTRACT FROM AUTHOR]

In this paper, the experimental studies of different flow density values have been carried on a designed and implemented flow meter being known as ultrasound working based on high frequency sound wave technology. The calibration of the flow meter has been realized using the results obtained from the study. The transit-time method has been used as the basis of the flow measure. For calibration of the flow meter, the experimental study has been carried on twenty different flow densities and the differences between the measured and the exact values have been observed. Then, to minimize these differences, two different calibration methods being 'linearequation' and 'artificial-neural-network' methods have been developed and the results have been obtained. Using the results, the two mentioned methods have been compared and those effects on the accuracy of the flow meter have been discussed. It has been shown that between these two methods the accuracy of the 'neuralnetwork' method is higher than that of the 'linear-equation' method. [ABSTRACT FROM AUTHOR]

In this paper, the Augmented Genetic Algorithm with Artificial Neural Network (AGANN) is expanded for optimization works, and its implementations to model problems are demonstrated. With the purpose of getting a faster algorithm, a neural network and a real coded genetic algorithm are hybridized in a new way. In this way, instead of predicting objective function calculation of a candidate, a properly trained neural network is used for predicting the candidate itself. At each step of the genetic process, using a simulated annealing based optimization procedure, the trained neural network produces an individual, which is a candidate solution of the optimization problem. Adding this candidate to the population at each step improves the exploration power of the genetic process. The proposed algorithm is tested for some test function problems. The results indicate that the computational efficiency of the implemented algorithm is tremendously high. Due to still being a genetic algorithm based technique, this method is also as robust as the pure genetic algorithms. [ABSTRACT FROM AUTHOR]

The marble cutting time which has a nonlinear structure changes depending on speed of circular saw rotation, forward speed and amount of cutting against time. Depending on this variables before mathematical modeling of the data optained from experiments made in laboratory environment are very difficult. In this study, intended to the completion time of marble cutting operation, based artificial neural network (ANN) modeling process were performed. 90% of the experimental data was given during training and 10% were used for testing purposes. According to the results obtained, performed the modelling was seen as applicable. [ABSTRACT FROM AUTHOR]

In this study, effect of the marble cutting parameters on energy consumption was estimated by using Artificial Neural Network and ANFIS. In both of these estimation methods, marble's hardness, speed of circular saw rotation and cutting speed parameters were used as inputs, specific energy factor, cutting energy in per unit volume was taken as output. The data for training networks were taken from a PC based block cutter machine (S/T) with circular diamond saw blade at Afyon Kocatepe University Labs. These data are cutting parameters of "Denizli Travertine" and "Bilecik Beige" type marbles. Some of these data were used for training and then all data were used for testing stage of the networks. To show the effectiveness of the study, estimated data were compared with experimental data. As conclusion, it was reported that estimated data were quite close to experimental result. [ABSTRACT FROM AUTHOR]