The development of technology generates huge amounts of non-textual information, such as images. An efficient image annotation and retrieval system is highly desired. Clustering algorithms make it possible to represent visual features of images with finite symbols. Based on this, many statistical models, which analyze correspondence between visual features and words and discover hidden semantics, have been published. These models improve the annotation and retrieval of large image databases. However, current state of the art including our previous work produces too many irrelevant keywords for images during annotation. In this paper, we propose a novel approach that augments the classical model with generic knowledge-based, WordNet. Our novel approach strives to prune irrelevant keywords by the usage of WordNet. To identify irrelevant keywords, we investigate various semantic similarity measures between keywords and finally fuse outcomes of all these measures together to make a final decision using Dempster-Shafer evidence combination. We have implemented various models to link visual tokens with keywords based on knowledge-based, WordNet and evaluated performance using precision, and recall using benchmark dataset. The results show that by augmenting knowledge-based with classical model we can improve annotation accuracy by removing irrelevant keywords.

Web prediction is a classification problem in which we attempt to predict the next set of Web pages that a user may visit based on the knowledge of the previously visited pages. Predicting user's behavior while serving the Internet can be applied effectively in various critical applications. Such application has traditional tradeoffs between modeling complexity and prediction accuracy. In this paper, we analyze and study Markov model and all- Kth Markov model in Web prediction. We propose a new modified Markov model to alleviate the issue of scalability in the number of paths. In addition, we present a new two-tier prediction framework that creates an example classifier EC, based on the training examples and the generated classifiers. We show that such framework can improve the prediction time without compromising prediction accuracy. We have used standard benchmark data sets to analyze, compare, and demonstrate the effectiveness of our techniques using variations of Markov models and association rule mining. Our experiments show the effectiveness of our modified Markov model in reducing the number of paths without compromising accuracy. Additionally, the results support our analysis conclusions that accuracy improves with higher orders of all- Kth model.

Predicting users' future requests in the World Wide Web can be applied effectively in many important applications, such as web search, latency reduction, and personalization systems. Such application has traditional tradeoffs between modeling complexity and prediction accuracy. In this paper, we study several hybrid models that combine different classification techniques, namely, Markov models, artificial neural networks (ANNs), and the All-Kth-Markov model, to resolve prediction using Dempster's rule. Such fusion overcomes the inability of the Markov model in predicting beyond the training data, as well as boosts the accuracy of ANN, particularly, when dealing with a large number of classes. We also employ a reduction technique, which uses domain knowledge, to reduce the number of classifiers to improve the predictive accuracy and the prediction time of ANNs. We demonstrate the effectiveness of our hybrid models by comparing our results with widely used techniques, namely, the Markov model, the All-Kth-Markov model, and association rule mining, based on a benchmark data set.