Software products are tools that find more uses in the society every day, both professionally and in daily lives of members of the society. This thesis focuses on the problems and possibilities related to current software products. By analysis of the problems and current issues in the software field, possible contribution of a design-oriented approach to software products is explored. The thesis is supported by a study in the form of a semi-structured observation.

This thesis presents a machine learning approach to the problem of opponent modeling in games of imperfect information. The efficiency of various artificial intelligence techniques are investigated in this domain. A sequential game is called imperfect information game if players do not have all the information about the current state of the game. A very popular example is the Texas Holdem Poker, which is used for realization of the suggested methods in this thesis. Opponent modeling is the system that enables a player to predict the behaviour of its opponent. In this study, opponent modeling problem is approached as a classification problem. An architecture with different classifiers for each phase of the game is suggested. Neural Networks, K-Nearest Neighbors (KNN) and Support Vector Machines are used as classifier. For modeling a particular player, KNN is found to be most successful amongst all, with a prediction accuracy of 88%. An ensemble learning system is proposed for modeling different playing styles and unknown ones. Computational complexity and parallelization of some calculations are also provided.

The goal of this thesis is to analyze and understand the risks of Chronic Obstructive Pulmonary Disease (COPD) and lung cancer with respect to the provinces of Turkey according to the results of spatial analysis. The insurance sector of the country needs that kind of analysis to make more precise pricing in insurance products. Especially in health and life insurance products, morbidities like COPD and lung cancer may aect the life expectancy as much as the premiums. COPD and lung cancer prevalence may exhibit spatial autocorrelation due to spatial similarity of provinces. Hence understanding of spatial pattern of COPD and lung cancer prevalence may provide better actuarial decisions. In this research, common risk factors of COPD and lung cancer are considered to be tobacco sales, air pollution, urbanization, gross schooling rate, life expectancy, median age and GDP per capita of the provinces. The spatial patterns of these factors in Turkey as well as their correlations to COPD and lung cancer prevalence are explored in this study. The raw data of the morbidities (COPD and lung cancer) are collected from the Social Seiv curity Institution (SGK) and the useful data are selected in these raw data. The data of the independent variables are collected and derived from the Turkish Statistical Institute (TUIK) and Tobacco and Alcohol Market Regulatory Authority (TAPDK). First of all, COPD prevalence ratios and lung cancer prevalence ratios are grouped by 81 provinces of Turkey and every morbidity is separated by gender. Then, it needs to be decided the variables which define prevalence of COPD and that of lung cancer. Age, gender, socio-economic status, urbanization, schooling rate, life expectancy, tobacco sales and air quality may be some of the random variables which are categorized by provinces for both morbidities. After data collection spatial analysis is applied with visualization, explanatory analysis and modeling by using Geographic Information Systems (GIS). In visualization, general spatial patterns are identified for morbidities and variables. In explanatory analysis part, proximity matrices are used to evaluate Moran&rsquo / s I values for understanding the spatial autocorrelation. Then, these Moran&rsquo / s I values are used for plotting correlograms in order to follow the spatial dependence better. After identifying spatial dependence of the variables, Ordinary Linear Regression and Spatial Regression models are established and compared. Finally, as a result of those findings in the analysis, actuarial risk assessments are found for both two morbidities with respect to provinces and gender. The risk assessments are mapped and compared with the explanatory variables in the models which are found in the previous part and the relations between risks and variables are observed. As a result, the parameters show spatial autocorrelation which means that / financial risk assessments of COPD and lung cancer should be taken into account when deciding the pricing of some actuarial products such as health insurance. Generally, spatial correlation is ignored in this kind of calculations, but due to the high autocorrelation the results may indicate serious change. From the actuarial perspective, the results of the analysis are suggested to be used in health insurance premium pricing. Since the analysis could not have been made on the basis of individuals, and financial burden of morbidities for insurance companies are not given clearly, it is not possible to calculate any health insurance product premium, but it is more appropriate to consider the importance of these risk results in the calculations of health insurance products.

This study deals with four location problems. The first problem is a brand new location problem on a line and considers the location decisions for depots and quarries in a highway construction project. We develop optimal solution properties of the problem. Using these properties, a dynamic programming algorithm is proposed. The second problem is also a brand new dynamic location problem on a line and locates concrete batching mobile and immobile facilities for a railroad construction project. We develop two mixed integer models to solve the problem. For solving large size problems, we propose a heuristic. Performances of models and the heuristic are tested on randomly generated instances plus a case study data and results are presented. The third problem is a generalization of the second problem to network locations. It is a dynamic version of the well known p-median problem and incorporates mobile facilities. The problem is to locate predetermined number of mobile and immobile facilities over a planning horizon such that sum of facility movement and allocation costs is minimized. Three constructive heuristics and a branch-and-price algorithm are proposed. Performances of these solution procedures are tested on randomly generated instances and results are presented. In the fourth problem we consider a special case of the third problem, allowing only conventional facilities. The algorithm for the third problem is improved so that generating columns and solving a mixed integer model are used repetitively. Performance of the algorithm is tested on randomly generated instances and results are presented.

We use the Garman-Kohlhagen model to compute the hedge and price of a participating forward contract on the US dollar that is written by a Turkish Bank. The algorithm is computed using actual market data and a weekly updated hedge is computed. We note that despite a weekly update and many assumptions made on the volatility and the interest rates the model gives a very reasonable hedge.

Automatic image annotation is the process of automatically producing words to de- scribe the content for a given image. It provides us with a natural means of semantic indexing for content based image retrieval. In this thesis, two novel automatic image annotation systems targeting di&amp / #64256 / erent types of annotated data are proposed. The &amp / #64257 / rst system, called Supervised Ensemble of Visual Descriptors (SEVD), is trained on a set of annotated images with prede&amp / #64257 / ned class labels. Then, the system auto- matically annotates an unknown sample depending on the classi&amp / #64257 / cation results. The second system, called Unsupervised Ensemble of Visual Descriptors (UEVD), assumes no class labels. Therefore, the annotation of an unknown sample is accomplished by unsupervised learning based on the visual similarity of images. The available auto- matic annotation systems in the literature mostly use a single set of features to train a single learning architecture. On the other hand, the proposed annotation systems utilize a novel model of image representation in which an image is represented with a variety of feature sets, spanning an almost complete visual information comprising color, shape, and texture characteristics. In both systems, a separate learning entity is trained for each feature set and these entities are gathered under an ensemble learning approach. Empirical results show that both SEVD and UEVD outperform some of the state-of-the-art automatic image annotation systems in equivalent experimental setups.

Credit risk is one of the major risks banks and financial institutions are faced with. With the New Basel Capital Accord, banks and financial institutions have the opportunity to improve their risk management process by using Internal Rating Based (IRB) approach. In this thesis, we focused on the internal credit rating process. First, a short overview of credit scoring techniques and validation techniques was given. By using real data set obtained from a Turkish bank about manufacturing firms, default prediction logistic regression, probit regression, discriminant analysis and classification and regression trees models were built. To improve the performances of the models the optimum sample for logistic regression was selected from the data set and taken as the model construction sample. In addition, also an information on how to convert continuous variables to ordered scaled variables to avoid difference in scale problem was given. After the models were built the performances of models for whole data set including both in sample and out of sample were evaluated with validation techniques suggested by Basel Committee. In most cases classification and regression trees model dominates the other techniques. After credit scoring models were constructed and evaluated, cut-off values used to map probability of default obtained from logistic regression to rating classes were determined with dual objective optimization. The cut-off values that gave the maximum area under ROC curve and minimum mean square error of regression tree was taken as the optimum threshold after 1000 simulation. Keywords: Credit Rating, Classification and Regression Trees, ROC curve, Pietra Index

In this thesis, we address the problem of real-time path search in partially observable grid worlds, and propose two single agent and one multi-agent search algorithm. The first algorithm, Real-Time Edge Follow (RTEF), is capable of detecting the closed directions around the agent by analyzing the nearby obstacles, thus avoiding dead-ends in order to reach a static target more effectively. We compared RTEF with a well-known algorithm, Real-Time A* (RTA*) proposed by Korf, and observed significant improvement. The second algorithm, Real-Time Moving Target Evaluation Search (MTES), is also able to detect the closed directions similar to RTEF, but in addition, determines the estimated best direction that leads to a static or moving target from a shorter path. Employing this new algorithm, we obtain an impressive improvement over RTEF with respect to path length, but at the cost of extra computation. We compared our algorithms with Moving Target Search (MTS) developed by Ishida and the off-line path planning algorithm A*, and observed that MTES performs significanlty better than MTS, and offers solutions very close to optimal ones produced by A*. Finally, we present Multi-Agent Real-Time Pursuit (MAPS) for multiple predators to capture a moving prey cooperatively. MAPS introduces two new coordination strategies namely Blocking Escape Directions (BES) and Using Alternative Proposals (UAL), which help the predators waylay the possible escape directions of the prey in coordination. We compared our coordination strategies with the uncoordinated one, and observed an impressive reduction in the number of moves to catch the prey.

Technological developments cause products to include more software which become a part of products and a determiner of product characteristics. Today&amp / #8217 / s way of life surrounded by technological developments caused users to engage in inevitable interaction with products. By means of that interaction, users&amp / #8217 / way of meeting their needs and attitudes towards products shaped in both functional and emotional dimensions. Although users&amp / #8217 / attitudes seem to be governing this process, the nature and the characteristics of the products affected the manner of this interaction. Consequently, the existence of the software as a product characteristic would affect the users&amp / #8217 / attitude towards products. This thesis focuses on the effects of the software existence on users&amp / #8217 / attitudes. The empirical study conducted in the thesis explored the effects of different levels of software existence in products. This empirical study analyzes three product groups and the results of that analysis indicated that the existence of software affects users&amp / #8217 / perceptions about products in a way that it causes the bond between the user and the owned product to get weaker in hedonic aspects.

A genetic algorithm using conventional heuristics as operators is considered in this study for the traveling salesman problem with backhauls (TSPB). Properties of a crossover operator (Nearest Neighbor Crossover, NNX) based on the nearest neighbor heuristic and the idea of using more than two parents are investigated in a series of experiments. Different parent selection and replacement strategies and generation of multiple children are tried as well. Conventional improvement heuristics are also used as mutation operators. It has been observed that 2-edge exchange and node insertion heuristics work well with NNX using only two parents. The best settings among different alternatives experimented are applied on traveling salesman problem with backhauls (TSPB). TSPB is a problem in which there are two groups of customers. The aim is to minimize the distance traveled visiting all the cities, where the second group can be visited only after all cities in the first group are already visited. The approach we propose shows very good performance on randomly generated TSPB instances.