Catálogo de publicaciones - libros

Prior studies on hotel website performance have primarily concentrated on frequency counting, content analysis or user behavioral approaches. These studies, however, failed to offer any insight that can accurately evaluate hotel website quality based on users’ assessment of attribute weights and performance ratings. This research proposes a fuzzy multicriteria analysis model which systematically integrates hotel guests’ preferences and fuzzy assessments of website attributes. The fuzzy evaluation of linguistic values by respondents offers a comprehensive approach for handling incomplete and imprecise user preferences to capture the realistic evaluation process. The research output will be a set of overall performance indices representing the cumulative effect of website features, which will offer a benchmark enabling hotels to evaluate their websites’ relative performance and ranking based on all relevant weighted attributes.

Ever since the advent of Internet, there has been an immense growth in the amount of image data that is available on the World Wide Web. With such a magnitude of image availability, an efficient and effective image retrieval system is required to make use of this information. This research presents an image matching and indexing technique that improvises on existing integrated image retrieval methods. The proposed system integrates query by topic and query by example specification methods. The topic-based image retrieval uses the structured format of HTML documents to retrieve relevant pages and potential match images. The query by example specification performs content-based image match for the retrieval of smaller and relatively closer results of the example image. The main goal is to develop a functional image search and indexing system without using a database and to demonstrate that better retrieval results can be achieved with this proposed hybrid search technique.

With the technological advances, the Internet has been an important part of everyday life. Governmental institutions and enterprises tend to advertise and market through the internet. With the travelling records of browsers, one can analyze the preference of web pages, further understand the demands of consumers, and promote the advertising and marketing. In this study, we use Maximum Forward Reference (MFR) algorithm to find the travel pattern of browsers from web logs. Simultaneously, experts are asked to evaluate the fuzzy importance weightings for different webs. Finally, we employ fuzzy data mining technique that combines apriori algorithm with fuzzy weights to determine the association rules. From the yielded association rules, one can be accurately aware of the information consumers need and which webs they prefer. This is important to governmental institutions and enterprises. Enterprises can find the commercial opportunities and improve the design of webs by means of this study. Governmental institutions can realize the needs of people from the obtained association rules, make the promotion of policy more efficiently, and provide better services.

A Web site generally contains a wide range of topics which provide information for users who have different access interests and goals. This information is not randomly scattered, but well organized under a hierarchy encoded in the hyperlink structure of a Web site. It is intended to mold the user’s mental models of how the information is organized. On the other hand, user traversals over hyperlinks between Web pages can reveal semantic relationships between these pages. Unfortunately, the link structure of a Web site which represent the Web designer’s expectation on visitors may be quite different from the organization expected by visitors to this site. Discovering the conceptual page hierarchy from a user’s angle can help web masters to have an sight into real relationships among the Web pages and refine the link structure of the Web site to facilitate effective user navigation. In this paper, we propose a method to generate a conceptual page hierarchy of a Web site on the basis of user traversal history. We use maximal forward references to model user’s traversal behavior over the underlying link hierarchy of a Web site. We then build a weighted directed graph to represent the inter-relationships between Web pages. Finally we apply a “” (MST) algorithm to generate a conceptual page hierarchy of the Web site. We demonstrate the effectiveness of our approach by conducting a preliminary experiment based on a real world Web data.

A novel approach is developed for Protein Secondary Structure Prediction based on Bayesian Neural Networks (BNN). BNN usually outperforms the traditional Back-Propagation Neural Networks (BPNN) due to its excellent ability to control the complexity of the model. Results indicates that BNN has an average overall three-state accuracy increase 3.65% and 4.01% on the 4-fold cross-validation data sets and TEST data set respectively, comparing with the traditional BPNN. Meanwhile, a so-called is presented, which will shorten the burn-in phase during the MCMC (Markov Chain Monte Carlo) simulation substantially.

In this paper we present a novel hybrid machine learning system for recognition of gene starts in human genome. The system makes predictions of gene start by extracting compositional features and CpG islands information from promoter regions. It combines a new promoter recognition model, coding theory, feature selection and dimensionality reduction with machine learning algorithm. Evaluation on Human chromosome 4, 21, 22 was 64.47% in sensitivity and 82.20% in specificity. Comparison with the three other systems revealed that our system had superior sensitivity and specificity in predicting gene starts. PromPredictor is written in MATLAB and requires Matlab to run. PromPredictor is freely available at www.whtelecom.com/Prompredictor.htm.

DNA microarray technology has demonstrated to be an effective methodology for the diagnosis of cancers by means of microarray data classification. Although much research has been conducted during the recent years to apply machine learning techniques for microarray data classification, there are two important issues that prevent the use of conventional machine learning techniques, namely the limited availability of training samples and the existence of various uncertainties (e.g. biological variability and experiment variability). This paper presents a new ensemble machine learning approach to address these issues in order to achieve a robust microarray data classification. Ensemble learning combines a set of base classifiers as a committee to make appropriate decisions when classifying new data instances. In order to enhance the performance of the ensemble learning process, the approach presented includes a procedure to select optimal ensemble members that maximize the behavioural diversity. The proposed approach has been verified by three microarray datasets for cancer diagnosis. Experimental results have demonstrated that the classifier constructed by the proposed method outperforms not only the classifiers generated by the conventional machine learning techniques, but also the classifiers generated by two widely-used conventional Bagging and Boosting ensemble learning methods.

Artificial Immune Systems are a new class of algorithms inspired by how the immune system recognizes, attacks and remembers intruders. This is a fascinating idea, but to be accepted for mainstream data mining applications, extensive benchmarking is needed to demonstrate the reliability and accuracy of these algorithms. In our research we focus on the AIRS classification algorithm. It has been claimed previously that AIRS consistently outperforms other algorithms. However, in these papers AIRS was compared to benchmark results from literature. To ensure consistent conditions we carried out benchmark tests on all algorithms using exactly the same set up. Our findings show that AIRS is a stable and robust classifier that produces around average results. This contrasts with earlier claims but shows AIRS is mature enough to be used for mainstream data mining.

It is well accepted that many real-life datasets are full of missing data. In this paper we introduce, analyze and compare several well known treatment methods for missing data handling and propose new methods based on Naive Bayesian classifier to estimate and replace missing data. We conduct extensive experiments on datasets from UCI to compare these methods. Finally we apply these models to a geriatric hospital dataset in order to assess their effectiveness on a real-life dataset.

In this paper, we design genetic algorithm and simulated annealing algorithm and their parallel versions to solve the Closest String Problem. Our implementation and experiments show usefulness of the parallel GA and SA algorithms.