Dissertations / Theses on the topic 'Attribute Similarity'

Software effort estimation by analogy is a viable alternative method to other estimation techniques, and in many cases, researchers found it outperformed other estimation methods in terms of accuracy and practitioners' acceptance. However, the overall performance of analogy based estimation depends on two major factors: similarity measure and attribute selection & weighting. Current similarity measures such as nearest neighborhood techniques have been criticized that have some inadequacies related to attributes relevancy, noise and uncertainty in addition to the problem of using categorical attributes. This research focuses on improving the efficiency and flexibility of analogy-based estimation to overcome the abovementioned inadequacies. Particularly, this thesis proposes two new approaches to model and handle uncertainty in similarity measurement method and most importantly to reflect the structure of dataset on similarity measurement using Fuzzy modeling based Fuzzy C-means algorithm. The first proposed approach called Fuzzy Grey Relational Analysis method employs combined techniques of Fuzzy set theory and Grey Relational Analysis to improve local and global similarity measure and tolerate imprecision associated with using different data types (Continuous and Categorical). The second proposed approach presents the use of Fuzzy numbers and its concepts to develop a practical yet efficient approach to support analogy-based systems especially at early phase of software development. Specifically, we propose a new similarity measure and adaptation technique based on Fuzzy numbers. We also propose a new attribute subset selection algorithm and attribute weighting technique based on the hypothesis of analogy-based estimation that assumes projects that are similar in terms of attribute value are also similar in terms of effort values, using row-wise Kendall rank correlation between similarity matrix based project effort values and similarity matrix based project attribute values. A literature review of related software engineering studies revealed that the existing attribute selection techniques (such as brute-force, heuristic algorithms) are restricted to the choice of performance indicators such as (Mean of Magnitude Relative Error and Prediction Performance Indicator) and computationally far more intensive. The proposed algorithms provide sound statistical basis and justification for their procedures. The performance figures of the proposed approaches have been evaluated using real industrial datasets. Results and conclusions from a series of comparative studies with conventional estimation by analogy approach using the available datasets are presented. The studies were also carried out to statistically investigate the significant differences between predictions generated by our approaches and those generated by the most popular techniques such as: conventional analogy estimation, neural network and stepwise regression. The results and conclusions indicate that the two proposed approaches have potential to deliver comparable, if not better, accuracy than the compared techniques. The results also found that Grey Relational Analysis tolerates the uncertainty associated with using different data types. As well as the original contributions within the thesis, a number of directions for further research are presented. Most chapters in this thesis have been disseminated in international journals and highly refereed conference proceedings.

Due to the increasing amount and complexity of data processed in real world applications, similarity search became a vital task to store and retrieve such data. However, missing attribute values are very frequent and metric access methods (MAMs), designed to support similarity search, do not operate on datasets when attribute values are missing. Currently, the approach to use the existing indexing techniques on datasets with missing attribute values just use an indicator to identify the missing values and employ a traditional indexing technique. Although, this approach can be applied over multidimensional indexing techniques, it is impractical for metric access methods. This dissertation presents the results of a research conducted to identify and deal with the issues related to indexing and querying datasets with missing values in metric spaces. An empirical analysis of the metric access methods when applied on incomplete datasets leads us to identify two main issues: distortion of the internal structure of the index when data are missing at random and skew of the index structure when data are not missing at random. Based on those findings, a new variant of the Slim-tree access method, called Hollow-tree, is presented. It employs new techniques that are capable to handle missing data issues when missingness is ignorable. The first technique includes a set of indexing policies that allow to index objects with missing attribute values and prevent distortions to occur in the internal structure of the indexes. The second technique targets the similarity queries to improve the query performance over incomplete datasets. This technique employs the fractal dimension of the dataset and the local density around the query object to estimate an ideal radius able to achieve an accurate query answer, considering data with missing values as a potential response. Results from experiments with a variety of real and synthetic datasets show that Hollow-tree achieves nearly 100% of precision and recall for Range queries and more than 90% for k Nearest Neighbor queries, while Slim-tree access method deteriorates with the increasing amount of missing values. The results confirm that the indexing technique helps to establish consistency in the index structure and the searching technique achieves a remarkable performance. When combined, the new techniques allow to explore properly all the available data even with high amounts of missing attribute values. As they are independent of the underlying access method, they can be adopted by a broad range of metric access methods, allowing to extend the class of MAMs.<br>O crescimento em quantidade e complexidade dos dados processados e armazenados torna a busca por similaridade uma tarefa fundamental para tratar esses dados. No entanto, atributos faltantes ocorrem freqüentemente, inviabilizando os métodos de acesso métricos (MAMs) projetados para apoiar a busca por similaridade. Assim, técnicas de tratamento de dados faltantes precisam ser desenvolvidas. A abordagem mais comum para executar as técnicas de indexação existentes sobre conjuntos de dados com valores faltantes é usar um indicador de valores faltantes e usar as técnicas de indexação tradicionais. Embora, esta técnica seja útil para os métodos de indexação multidimensionais, é impraticável para os métodos de acesso métricos. Esta dissertação apresenta os resultados da pesquisa realizada para identificar e lidar com os problemas de indexação e recuperação de dados em espaços métricos com valores faltantes. Uma análise experimental dos MAMs aplicados a conjuntos de dados incompletos identificou dois problemas principais: distorção na estrutura interna do índice quando a falta é aleatória e busca tendenciosa na estrutura do índice quando o processo de falta não é aleatório. Uma variante do MAM Slim-tree, chamada Hollow-tree foi proposta com base nestes resultados. A Hollow-tree usa novas técnicas de indexação e de recuperação de dados com valores faltantes quando o processo de falta é aleatório. A técnica de indexação inclui um conjunto de políticas de indexação que visam a evitar distorções na estrutura interna dos índices. A técnica de recuperação de dados melhora o desempenho das consultas por similaridade sobre bases de dados incompletas. Essas técnicas utilizam o conceito de dimensão fractal do conjunto de dados e a densidade local da região de busca para estimar um raio de busca ideal para obter uma resposta mais correta, considerando os dados com valores faltantes como uma resposta potencial. As técnicas propostas foram avaliadas sobre diversos conjuntos de dados reais e sintéticos. Os resultados mostram que a Hollow-tree atinge quase 100% de precisão e revocação para consultas por abrangência e mais de 90% para k vizinhos mais próximos, enquanto a Slim-tree rapidamente deteriora com o aumento da quantidade de valores faltantes. Tais resultados indicam que a técnica de indexação proposta ajuda a estabelecer a consistência na estrutura do índice e a técnica de busca pode ser realizada com um desempenho notável. As técnicas propostas são independentes do MAM básico usado e podem ser aplicadas em uma grande variedade deles, permitindo estender a classe dos MAMs em geral para tratar dados faltantes.

Um dos grandes desafios em mineração de dados é a integração de dados temporais ao seu processo. Existe um grande número de aplicações emergentes que envolvem dados temporais, incluindo a identificação de transações fraudulentas em cartões de crédito e ligações telefônicas, a detecção de intrusão em sistemas computacionais, a predição de estruturas secundárias de proteínas, a análise de dados provenientes de sensores, entre muitas outras. Neste trabalho, tem-se interesse na classificação de séries temporais que representam sinais de áudio. Como aplicação principal, tem-se interesse em classificar sinais de insetos coletados por um sensor óptico, que deve ser capaz de contar e classificar os insetos de maneira automática. Apesar de serem coletados opticamente, os sinais capturados se assemelham a sinais de áudio. O objetivo desta pesquisa é comparar métodos de classificação por similaridade e por extração de atributos que possam ser utilizados no contexto da classificação de insetos. Para isso, foram empregados os principais métodos de classificação de sinais de áudio, que têm sido propostos para problemas como reconhecimento de instrumentos musicais, fala e espécies animais. Neste trabalho, é mostrado que, de modo geral, a abordagem por extração de atributos é mais eficaz do que a classificação por similaridade. Mais especificamente, os melhores resultados são obtidos com a utilização de coeficientes mel-cepstrais. Este trabalho apresenta contribuições significativas em outras aplicações, também relacionadas à análise de séries temporais e sinais de áudio, por similaridade e por extração de atributos<br>One of the major challenges in data mining is the integration of temporal data to its process. There are a number of emerging applications that involve temporal data, including fraud detection in credit card transactions and phone calls, intrusion detection in computer systems, the prediction of secondary structures of proteins, the analysis of data from sensors, and many others. In this work, our main interest is the classification of time series that represent audio signals. Our main interest is an application for classifying signals of insects collected from an optical sensor, which should count and classify insects automatically. Although these signals are optically collected, they resemble audio signals. The objective of this research is to compare classification methods based on similarity and feature extraction in the context of insects classification. For this purpose, we used the main classification methods for audio signals, which have been proposed for problems such as musical instrument, speech and animal species recognition. This work shows that, in general, the approach based on feature extraction is more accurate than the classification by similarity. More specifically, the best results are obtained with mel-frequency cepstrum coefficients. This work also presents significant contributions in other applications, also related to the analysis of time series and audio signals by similarity and feature extraction

This thesis makes an original contribution to knowledge in the fi eld of data objects' comparison where the objects are described by attributes of fuzzy or heterogeneous (numeric and symbolic) data types. Many real world database systems and applications require information management components that provide support for managing such imperfect and heterogeneous data objects. For example, with new online information made available from various sources, in semi-structured, structured or unstructured representations, new information usage and search algorithms must consider where such data collections may contain objects/records with di fferent types of data: fuzzy, numerical and categorical for the same attributes. New approaches of similarity have been presented in this research to support such data comparison. A generalisation of both geometric and set theoretical similarity models has enabled propose new similarity measures presented in this thesis, to handle the vagueness (fuzzy data type) within data objects. A framework of new and unif ied similarity measures for comparing heterogeneous objects described by numerical, categorical and fuzzy attributes has also been introduced. Examples are used to illustrate, compare and discuss the applications and e fficiency of the proposed approaches to heterogeneous data comparison.<br>Libyan Embassy

This thesis makes an original contribution to knowledge in the fi eld of data objects' comparison where the objects are described by attributes of fuzzy or heterogeneous (numeric and symbolic) data types. Many real world database systems and applications require information management components that provide support for managing such imperfect and heterogeneous data objects. For example, with new online information made available from various sources, in semi-structured, structured or unstructured representations, new information usage and search algorithms must consider where such data collections may contain objects/records with di fferent types of data: fuzzy, numerical and categorical for the same attributes. New approaches of similarity have been presented in this research to support such data comparison. A generalisation of both geometric and set theoretical similarity models has enabled propose new similarity measures presented in this thesis, to handle the vagueness (fuzzy data type) within data objects. A framework of new and unif ied similarity measures for comparing heterogeneous objects described by numerical, categorical and fuzzy attributes has also been introduced. Examples are used to illustrate, compare and discuss the applications and e fficiency of the proposed approaches to heterogeneous data comparison.

Cette dernière décennie, la modélisation des villes 3D est devenue l'un des enjeux de la recherche multimédia et un axe important en reconnaissance d'objets. Dans cette thèse nous nous sommes intéressés à localiser différentes primitives, plus particulièrement les fenêtres, dans les façades de Paris. Dans un premier temps, nous présentons une analyse des façades et des différentes propriétés des fenêtres. Nous en déduisons et proposons ensuite un algorithme capable d'extraire automatiquement des hypothèses de fenêtres. Dans une deuxième partie, nous abordons l'extraction et la reconnaissance des primitives à l'aide d'appariement de graphes de contours. En effet une image de contours est lisible par l'oeil humain qui effectue un groupement perceptuel et distingue les entités présentes dans la scène. C'est ce mécanisme que nous avons cherché à reproduire. L'image est représentée sous la forme d'un graphe d'adjacence de segments de contours, valué par des informations d'orientation et de proximité des segments de contours. Pour la mise en correspondance inexacte des graphes, nous proposons plusieurs variantes d'une nouvelle similarité basée sur des ensembles de chemins tracés sur les graphes, capables d'effectuer les groupements des contours et robustes aux changements d'échelle. La similarité entre chemins prend en compte la similarité des ensembles de segments de contours et la similarité des régions définies par ces chemins. La sélection des images d'une base contenant un objet particulier s'effectue à l'aide d'un classifieur SVM ou kppv. La localisation des objets dans l'image utilise un système de vote à partir des chemins sélectionnés par l'algorithme d'appariement<br>This last decade, modeling of 3D city became one of the challenges of multimedia search and an important focus in object recognition. In this thesis we are interested to locate various primitive, especially the windows, in the facades of Paris. At first, we present an analysis of the facades and windows properties. Then we propose an algorithm able to extract automatically window candidates. In a second part, we discuss about extraction and recognition primitives using graph matching of contours. Indeed an image of contours is readable by the human eye, which uses perceptual grouping and makes distinction between entities present in the scene. It is this mechanism that we have tried to replicate. The image is represented as a graph of adjacency of segments of contours, valued by information orientation and proximity to edge segments. For the inexact matching of graphs, we propose several variants of a new similarity based on sets of paths, able to group several contours and robust to scale changes. The similarity between paths takes into account the similarity of sets of segments of contours and the similarity of the regions defined by these paths. The selection of images from a database containing a particular object is done using a KNN or SVM classifier

This thesis proposes a novel directed hypergraph based model for any database. We introduce the notion of association rules for multi-valued attributes, which is an adaptation of the definition of quantitative association rules known in the literature. The association rules for multi-valued attributes are integrated in building the directed hypergraph model. This model allows to capture attribute-level associations and their strength. Basing on this model, we provide association-based similarity notions between any two attributes and present a method for finding clusters of similar attributes. We then propose algorithms to identify a subset of attributes known as a leading indicator that influences the values of almost all other attributes. Finally, we present an association-based classifier that can be used to predict values of attributes. We demonstrate the effectiveness of our proposed model, notions, algorithms, and classifier through experiments on a financial time-series data set (S&P 500).

<p>This study examines how consumers perceive the overall quality and similarity/fit of the brand extensions (Nike camera, Nike socks, and Nike golf balls) in relation to the original brand (Nike athletic shoes) and how these perceptions influence their attitudes towards the extensions. The researcher proposes that the perceived overall quality of the brand extensions will be congruent to that of the original brand and that the attitude towards the brand extensions will be favourable only when there is a perceived similarity/fit between the brand extensions and the original brand. In order to get these insights, the study surveyed 147 undergraduate and postgraduate students from the University of the Western Cape across all faculties. A non-probability convenience sampling method was used to access respondents. To collect data, qualitative and quantitative methods were employed using a questionnaire which consisted of open-ended and closed-ended questions in the form of free associations technique and 5-point Likert scales. The data gathered was analysed by means of descriptive statistics and Spearman correlation coefficient. The findings indicate two things. (1) respondents only perceived Nike socks&rsquo<br>overall quality to be congruent to the original brand, Nike athletic shoes. (2) respondents only perceived Nike socks to be strongly similar to the original brand, Nike athletic shoes.</p>

Cette dernière décennie, la modélisation des villes 3D est devenue l'un des enjeux de la recherche multimédia et un axe important en reconnaissance d'objets. Dans cette thèse nous nous sommes intéressés à localiser différentes primitives, plus particulièrement les fenêtres, dans les façades de Paris. Dans un premier temps, nous présentons une analyse des façades et des différentes propriétés des fenêtres. Nous en déduisons et proposons ensuite un algorithme capable d'extraire automatiquement des hypothèses de fenêtres. Dans une deuxième partie, nous abordons l'extraction et la reconnaissance des primitives à l'aide d'appariement de graphes de contours. En effet une image de contours est lisible par l'oeil humain qui effectue un groupement perceptuel et distingue les entités présentes dans la scène. C'est ce mécanisme que nous avons cherché à reproduire. L'image est représentée sous la forme d'un graphe d'adjacence de segments de contours, valué par des informations d'orientation et de proximité des segments de contours. Pour la mise en correspondance inexacte des graphes, nous proposons plusieurs variantes d'une nouvelle similarité basée sur des ensembles de chemins tracés sur les graphes, capables d'effectuer les groupements des contours et robustes aux changements d'échelle. La similarité entre chemins prend en compte la similarité des ensembles de segments de contours et la similarité des régions définies par ces chemins. La sélection des images d'une base contenant un objet particulier s'effectue à l'aide d'un classifieur SVM ou kppv. La localisation des objets dans l'image utilise un système de vote à partir des chemins sélectionnés par l'algorithme d'appariement.

碩士<br>真理大學<br>觀光事業學系碩士班<br>107<br>This study aims to investigate the perceived importance and satisfaction of users on bikeways of Tamsui Golden Riverside bikeway and Bali Left Bank bike path prioritize the improvement strategies to enhance bikeway service qualities and further increase user’s satisfaction by using Similarity-Based diagonal Importance-Performance Analysis. The results of the study, show found that the service attributes of Tamsui Golden Riverside bikeway include fresh air, bikeway cleanliness, width, slope, curvature, passing safety and pavement flatness need to be improved immediately. The service attributes needed to be improved immediately of "Bali Left Bank bike path" are fresh air, Bikeway cleanliness, length, width, slope, curvature, passing safety, pavement flatness, traffic signs, rest areas and toilet.

In this study, deductive reasoning with conditional statements of the form IF x THEN y was investigated in a transfer of training paradigm. Subjects induced conditional rules under different training conditions and then solved a series of deductive syllogisms based on conditional rules the content of which was either related (near-transfer) or unrelated (far-transfer) to that of the previously induced rules. The conditions under which the conditional rules were induced varied in terms of the acquisition mode (prediction/diagnosis) and in terms of the size of the attribute dimensions from which rule instances were drawn (binary/trinary). The deductive reasoning task, as the criterion task for transfer, included all four types of conditional syllogisms (Modus Ponens, MP; Denying the Antecedent, DA, Affirming the Consequent, AC, and Modus Tollens, MT). The accuracy of deductive performance on near-transfer was significantly higher than response accuracy on far-transfer on the DA & AC but not on the MP & MT argument forms. Transfer of conditional rule knowledge induced in a predictive mode was significantly better compared to a diagnostic mode on the AC form. Overall, near as well as far-transfer performance was similar for the binary and trinary conditions. In addition, neither the binary nor the trinary condition resulted in improved far-transfer performance compared to controls. The results provide evidence against purely instance-based as well as purely rule-based models of deductive reasoning. Instead, they suggest that conditional knowledge is represented at an intermediate level of abstraction, and that conditional reasoning is adaptive to the form of the deductive argument. Different types of deductive argument types may trigger different reasoning strategies.

碩士<br>國立臺北大學<br>資訊管理研究所<br>93<br>Due to the rapid development of e-Commerce, the Internet users can search product information easy through the online store in accordance with their own interests. When the users browse the product online, the inappropriate sort make the users hard to find the target goods they need. How to classify the products efficiently and properly is now an important issue for online store managers. This study proposed an automatic product classification mechanism classify products according to similarities. The similarities are calculated according to products’ specifics. This mechanism assigns all products into the proper groups. The online store websites may display products according to this obtain groups data and similarities of all products. Once a new product record is inserted into the database, all products in the database including the new one will be re-classified according to the automatic product classification mechanism this study proposed, and the similarities and groups would also be re-new. This can make products classification automatically. In the traditional classification methods, the users have to leave their personal information and use an account/password if the websites hope to sort the product messages according to customers’ preference. Nowadays, the Internet user pays more attention on privacy and information security, consumers may be unwilling to leave their personal information. However, the automatic product classification mechanism this study proposed let users could choice the product classification they prefer without providing personal information. Users pick one of the representative product items and then all the other products would sort according to the similarities of products to this selected item. Since that the products are sort according to consumers’ preference, the purchase rate might be raised. With this automatic merchandise clustering system, online store managers would not be worried about how to classify and how to sort the new added products. All they need to do is to put all the specifications of the new product into the database correctly and the automatically classification mechanism would do all the others.

碩士<br>朝陽科技大學<br>資訊管理系碩士班<br>90<br>Image similarity retrieval is a very important issue. An efficient similarity retrieval algorithm is needed when developing a multimedia database. In this dissertation, we propose some efficient techniques to solve image similarity retrieval problem. A user can pose a Content-Based query by specifying a query image example. First, the concept of image partition is applied to solve the color distribution problem. For a single object on monotonous background images, the Majority function is proposed to extract features in image and use the threshold r to adjust the similar precision. For the ordinal images, a new scheme that called “Absorber” is designed to extract the feature of an image and according the results to construct the feature table. Besides, the similar order is sorted along the distance between the query image and the image index in database.

No<br>Real-world data collections are often heterogeneous (represented by a set of mixed attributes data types: numerical, categorical and fuzzy); since most available similarity measures can only be applied to one type of data, it becomes essential to construct an appropriate similarity measure for comparing such complex data. In this paper, a framework of new and unified similarity measures is proposed for comparing heterogeneous objects described by numerical, categorical and fuzzy attributes. Examples are used to illustrate, compare and discuss the applications and efficiency of the proposed approach to heterogeneous data comparison and clustering.

Cette dernière décennie, la modélisation des villes 3D est devenue l'un des enjeux de la recherche multimédia et un axe important en reconnaissance d'objets. Dans cette thèse nous nous sommes intéressés à localiser différentes primitives, plus particulièrement les fenêtres, dans les façades de Paris. Dans un premier temps, nous présentons une analyse des façades et des différentes propriétés des fenêtres. Nous en déduisons et proposons ensuite un algorithme capable d'extraire automatiquement des hypothèses de fenêtres. Dans une deuxième partie, nous abordons l'extraction et la reconnaissance des primitives à l'aide d'appariement de graphes de contours. En effet une image de contours est lisible par l'oeil humain qui effectue un groupement perceptuel et distingue les entités présentes dans la scène. C'est ce mécanisme que nous avons cherché à reproduire. L'image est représentée sous la forme d'un graphe d'adjacence de segments de contours, valué par des informations d'orientation et de proximité des segments de contours. Pour la mise en correspondance inexacte des graphes, nous proposons plusieurs variantes d'une nouvelle similarité basée sur des ensembles de chemins tracés sur les graphes, capables d'effectuer les groupements des contours et robustes aux changements d'échelle. La similarité entre chemins prend en compte la similarité des ensembles de segments de contours et la similarité des régions définies par ces chemins. La sélection des images d'une base contenant un objet particulier s'effectue à l'aide d'un classifieur SVM ou kppv. La localisation des objets dans l'image utilise un système de vote à partir des chemins sélectionnés par l'algorithme d'appariement<br>This last decade, modeling of 3D city became one of the challenges of multimedia search and an important focus in object recognition. In this thesis we are interested to locate various primitive, especially the windows, in the facades of Paris. At first, we present an analysis of the facades and windows properties. Then we propose an algorithm able to extract automatically window candidates. In a second part, we discuss about extraction and recognition primitives using graph matching of contours. Indeed an image of contours is readable by the human eye, which uses perceptual grouping and makes distinction between entities present in the scene. It is this mechanism that we have tried to replicate. The image is represented as a graph of adjacency of segments of contours, valued by information orientation and proximity to edge segments. For the inexact matching of graphs, we propose several variants of a new similarity based on sets of paths, able to group several contours and robust to scale changes. The similarity between paths takes into account the similarity of sets of segments of contours and the similarity of the regions defined by these paths. The selection of images from a database containing a particular object is done using a KNN or SVM classifier

碩士<br>長庚大學<br>企業管理研究所<br>98<br>In Multiple Attribute Decision Making (MADM), how to assess the attribute weight properly is very important. Because different weight distribution would cause totally different decision result. In recent years, the IFS was applied to solve MADM problems for getting accurate information, but it makes our data and decision matrix get more complicated and contain more uncertainty. Therefore, it is such an important issue to make sure of the credibility of data and make correct judgement by itself. However, the research on MADM with the credibility of data is little in the past. In our research, we test Entropy Method, which is an objective weight method that being utilized most under intuitionistic fuzzy environment. We utilize the nature of IF entropy to assess the attribute weight based on the distance and similarity measures instead of standard step of information. According to the experiment result, we discover that there has no effect on the weight value among different measures although it seems complex between two formulas. Besides, we always think more complicated decision matrix would cause larger different result, because of the experiment, we understand this perspective is not correct. There has entirely different situation in the decision matrix which is composed of some specific attributes and alternatives, even in the small size.

碩士<br>南臺科技大學<br>企業管理系<br>106<br>This study mainly discusses whether the degree of business familization and the personality traits of the founder influence the personality preference of the company during personnel selection in the retail industry. This study takes manager of retail industry as the research object, adopts a questionnaire survey method, and results 208 valid questionnaires. SPSS statistical software is used to analyze the questionnaire data. Findings of the study are: 1. Managers of retail industry prefer employees with agreeableness, conscientiousness and emotional sensitive during personnel selection. 2. The degree of business familization of the retail enterprise has no significant effect on the personality traits preferences during personnel selection. 3. The personality traits of the founders have significant impacts on personality preferences of corporate during personnel selection. Finally, this paper provides some suggestions for personnel selection in the future.

Music services, media players and managers provide support for content classification and access based on filtering metadata values, statistics of access and user ratings. This approach fails to capture characteristics of mood and personal history that are often the deciding factors when creating personal playlists and collections in music. This dissertation work presents MusicWiz, a music management environment that combines traditional metadata with spatial hypertext-based expression and automatically extracted characteristics of music to generate personalized associations among songs. MusicWiz’s similarity inference engine combines the personal expression in the workspace with assessments of similarity based on the artists, other metadata, lyrics and the audio signal to make suggestions and to generate playlists. An evaluation of MusicWiz with and without the workspace and suggestion capabilities showed significant differences for organizing and playlist creation tasks. The workspace features were more valuable for organizing tasks, while the suggestion features had more value for playlist creation activities.

博士<br>國立成功大學<br>工業與資訊管理學系<br>102<br>The importance of small dataset learning problems has arisen in past decades because of the short product lifecycles caused by the increasing pressure of global competition. Although statistical approaches and the machine learning algorithms are widely applied to extract information from data, these are basically developed based on the assumption that training samples can afford to represent the whole population properties. Consequently, as the properties the training samples contain are limited, the knowledge that the algorithms extract may be confined. The virtual sample generation approaches, taken as one kind of data pretreatment methods, have proved their effectiveness when handling small datasets. Further, uniquely considering the occurrence relationship between attributes in the value generation procedure, this research proposes a non-parametric process to learn the trend similarities between attributes, and then based on which to estimate the corresponding ranges that attribute values may locate when other attributes’ value are given. Through the triangular membership functions which represent the attribute sample distributions, stepwise estimating the ranges for attribute values generation between, virtual samples are then formed. In the experiment, three real cases taken from related works are examined with the modeling tools including the M5' model tree, the multiple linear regression, the support vector regression, and the back-propagation neural network. The results show that the forecasting accuracies of the four modeling tools are improved when training sets contain virtual samples. In addition, the outcomes of the proposed procedure show significantly lower predictive errors than those of other approaches.