Large systems often have a long life-span and their system and software architecture design comprise many intricately related elements. The verification and maintenance of these architecture designs require an understanding of how and why the system are constructed. Design rationale is the reasoning behind a design and it provides an explanation of the design. However, the reasoning is often undocumented or unstructured in practice. This causes difficulties in the understanding of the original design, and makes it hard to detect inconsistencies, omissions and conflicts without any explanations to the intricacies of the design. Research into design rationale in the past has focused on argumentation-based design deliberations. Argumentation-based design rationale models provide an explicit representation of design rationale. However, these methods are ineffective in communicating design reasoning in practice because they do not support tracing to design elements and requirements in an effective manner. In this thesis, we firstly report a survey of practising architects to understand their perception of the value of design rationale and how they use and document this knowledge. From the survey, we have discovered that practitioners recognize the importance of documenting design rationale and frequently use them to reason about their design choices. However, they have indicated certain barriers to the use and documentation of design rationale. The results have indicated that there is no systematic approach to using and capturing design rationale in current architecture design practice. Using these findings, we address the issues of representing and applying architecture design rationale. We have constructed a rationale-based architecture model to represent design rationale, design objects and their relationships, which we call Architecture Rationale and Element Linkage (AREL). AREL captures both qualitative and quantitative rationale for architecture design. Quantitative rationale uses costs, benefits and risks to justify architecture decisions. Qualitative rationale documents the issues, arguments, alternatives and tradeoffs of a design decision. With the quantitative and qualitative rationale, the AREL model provides reasoning support to explain why architecture elements exist and what assumptions and constraints they depend on. Using a causal relationship in the AREL model, architecture decisions and architecture elements are linked together to explain the reasoning of the architecture design. Architecture Rationalisation Method (ARM) is a methodology that makes use of AREL to facilitate architecture design. ARM uses cost, benefit and risk as fundamental elements to rank and compare alternative solutions in the decision making process. Using the AREL model, we have proposed traceability and probabilistic techniques based on Bayesian Belief Networks (BBN) to support architecture understanding and maintenance. These techniques can help to carry out change impact analysis and rootcause analysis. The traceability techniques comprise of forward, backward and evolution tracings. Architects can trace the architecture design to discover the change impacts by analysing the qualitative reasons and the relationships in the architecture design. We have integrated BBN to AREL to provide an additional method where probability is used to evaluate and reason about the change impacts in the architecture design. This integration provides quantifiable support to AREL to perform predictive, diagnostic and combined reasoning. In order to align closely with industry practices, we have chosen to represent the rationale-based architecture model in UML. In a case study, the AREL model is applied retrospectively to a real-life bank payment systems to demonstrate its features and applications. Practising architects who are experts in the electronic payment system domain have been invited to evaluate the case study. They have found that AREL is useful in helping them understand the system architecture when they compared AREL with traditional design specifications. They have commented that AREL can be useful to support the verification and maintenance of the architecture because architects do not need to reconstruct or second-guess the design reasoning. We have implemented an AREL tool-set that is comprised of commercially available and custom-developed programs. It enables the capture of architecture design and its design rationale using a commercially available UML tool. It checks the well-formedness of an AREL model. It integrates a commercially available BBN tool to reason about the architecture design and to estimate its change impacts.

The need of today�s society for new technology has resulted in the development of a growing number of software systems. Developing a software system is a complex endeavour that requires a large amount of time. This amount of time is referred to as software development effort. Software development effort is the sum of hours spent by all individuals involved. Therefore, it is not equal to the duration of the development. Accurate prediction of the effort at an early stage of development is an important factor in the successful completion of a software system, since it enables the developing organization to allocate and manage their resource effectively. However, for many software systems, accurately predicting the effort is a challenge. Hence, a model that assists in the prediction is of active interest to software practitioners and researchers alike. Software development effort varies depending on many variables that are specific to the system, its developmental environment and the organization in which it is being developed. An accurate model for predicting software development effort can often be built specifically for the target system and its developmental environment. A local dataset of similar systems to the target system, developed in a similar environment, is then used to calibrate the model. However, such a dataset often consists of fewer than 10 software systems, causing a serious problem in the prediction, since predictive accuracy of existing models deteriorates as the size of the dataset decreases. This research addressed this problem with a new approach using Bayesian statistics. This particular approach was chosen, since the predictive accuracy of a Bayesian statistical model is not so dependent on a large dataset as other models. As the size of the dataset decreases to fewer than 10 software systems, the accuracy deterioration of the model is expected to be less than that of existing models. The Bayesian statistical model can also provide additional information useful for predicting software development effort, because it is also capable of selecting important variables from multiple candidates. In addition, it is parametric and produces an uncertainty estimate. This research developed new Bayesian statistical models for predicting software development effort. Their predictive accuracy was then evaluated in four case studies using different datasets, and compared with other models applicable to the same small dataset. The results have confirmed that the best new models are not only accurate but also consistently more accurate than their regression counterpart, when calibrated with fewer than 10 systems. They can thus replace the regression model when using small datasets. Furthermore, one case study has shown that the best new models are more accurate than a simple model that predicts the effort by calculating the average value of the calibration data. Two case studies has also indicated that the best new models can be more accurate for some software systems than a case-based reasoning model. Since the case studies provided sufficient empirical evidence that the new models are generally more accurate than existing models compared, in the case of small datasets, this research has produced a methodology for predicting software development effort using the new models.

Thesis (Ph. D.)--University of Hong Kong, 2007. / Title proper from title frame. Also available in printed format.

Thesis (Ph.D.)--Utrecht University, 2006. / Includes bibliographical references (p. [133]-137).

<p>Given a random sample from a multivariate probability distribution <em>p</em>, the maximum minimum parents and children algorithm locates the skeleton of the directed acyclic graph of a Bayesian network for <em>p</em> provided that there exists a faithful Bayesian network and that the dependence structure derived from data is the same as that of the underlying probability distribution.</p><p>The aim of this thesis is to examine the consequences when one of these conditions is not fulfilled. There are some circumstances where the algorithm works well even if there does not exist a faithful Bayesian network, but there are others where the algorithm fails.</p><p>The MMPC tests for conditional independence between the variables and assumes that if conditional independence is not rejected, then the conditional independence statement holds. There are situations where this procedure leads to conditional independence being accepted that contradict conditional dependence relations in the data. This leads to edges being removed from the skeleton that are necessary for representing the dependence structure of the data.</p>

<p>The main idea of our work is the calibration parameters for the Heston stochastic volatility model. We make this procedure by using the OMXS30 index from the NASDAQ OMX Nordic Exchange Market. We separate our data into the stable period and high-volatility period on this Nordic Market. Deviation detection problem are solved using the Bayesian analysis of change-points. We estimate parameters of the Heston model for each of periods and make some conclusions.</p>

Logic Sampling, Likelihood Weighting and AIS-BN are three variants of stochastic sampling, one class of approximate inference for Bayesian networks. We summarize the ideas underlying each algorithm and the relationship among them. The results from a set of empirical experiments comparing Logic Sampling, Likelihood Weighting and AIS-BN are presented. We also test the impact of each of the proposed heuristics and learning method separately and in combination in order to give a deeper look into AIS-BN, and see how the heuristics and learning method contribute to the power of the algorithm. Key words: belief network, probability inference, Logic Sampling, Likelihood Weighting, Importance Sampling, Adaptive Importance Sampling Algorithm for Evidential Reasoning in Large Bayesian Networks(AIS-BN), Mean Percentage Error (MPE), Mean Square Error (MSE), Convergence Rate, heuristic, learning method. / Graduation date: 2002

Bayesian networks are used for building intelligent agents that act under uncertainty. They are a compact representation of agents' probabilistic knowledge. A Bayesian network can be viewed as representing a factorization of a full joint probability distribution into the multiplication of a set of conditional probability distributions. Independence of causal influence enables one to further factorize the conditional probability distributions into a combination of even smaller factors. The efficiency of inference in Bayesian networks depends on how these factors are combined. Finding an optimal combination is NP-hard. We propose a new method for efficient inference in large Bayesian networks, which is a combination of new representations and new combination algorithms. We present new, purely multiplicative representations of independence of causal influence models. They are easy to use because any standard inference algorithm can work with them. Also, they allow for exploiting independence of causal influence fully because they do not impose any constraints on combination ordering. We develop combination algorithms that work with heuristics. Heuristics are generated automatically by using machine learning techniques. Empirical studies, based on the CPCS network for medical diagnosis, show that this method is more efficient and allows for inference in larger networks than existing methods. / Graduation date: 1999

This thesis describes the application of Bayesian networks for monitoring and diagnosis of a multi-stage manufacturing process, specifically a high speed production part at Hewlett Packard. Bayesian network "part models" were designed to represent individual parts in-process. These were combined to form a "process model", which is a Bayesian network model of the entire manufacturing process. An efficient procedure is designed for managing the "process network". Simulated data is used to test the validity of diagnosis made from this method. In addition, a critical analysis of this method is given, including computation speed concerns, accuracy of results, and ease of implementation. Finally, a discussion on future research in the area is given. / Graduation date: 1999

Graduation date: 1994