No abstract

One of the successful methods in classification problems is feature selection. Feature selection algorithms; try to classify an instance with lower dimension, instead of huge number of required features, with higher and acceptable accuracy. In fact an instance may contain useless features which might result to misclassification. An appropriate feature selection methods tries to increase the effect of significant features while ignores insignificant subset of features. In this work feature selection is formulated as an optimization problem and a novel feature selection procedure in order to achieve to a better classification results is proposed. Experiments over a standard benchmark demonstrate that applying Bee Colony Optimization in the context of feature selection is a feasible approach and improves the classification results.

No abstract

Many efforts have been done to tackle the problem of information abundance in the World Wide Web. Growth in the number of web users and the necessity of making the information available on the web make web recommender systems very critical and popular. Recommender systems use the knowledge obtained through the analysis of users' navigational behavior to customize a web site to the needs of each particular user or set of users. Most of the existing recommender systems use either content-based or collaborative filtering approach. It is difficult to decide which one of these approaches is the most effective one to be used, as each of them has both strengths and weaknesses. Therefore, a combination of different methods as a hybrid system can overcome these limitations and increase the effectiveness of the system. This paper introduces a new hybrid recommender system by exploiting a combination of collaborative filtering and content-based approaches in a way that resolves the drawbacks of each approach and makes a great improvement in the variety of recommendations in comparison to each individual approach. We introduce a new fuzzy clustering approach based on genetic algorithms and create a two-layer graph. After applying this clustering algorithm to both layers of the graph, we compute the similarity between web pages and users, and propose recommendations using the content-based, collaborative and hybrid approaches. A detailed comparison on all the mentioned approaches shows that the hybrid approach recommends the web pages which haven't been yet viewed by any user, more accurately and precisely than other approaches. Therefore, the evaluation of the results reveals that the novel proposed combination approach achieves more accurate predictions and more appropriate recommendations than each individual approach.

One of the major challenges in Web search pertains to the correct interpretation of users' intent. Query Expansion is one of the well-known approaches for determining the intent of the user by addressing the vocabulary mismatch problem. A limitation of the current query expansion approaches is that the relations between the query words and the expanded words is limited. In this thesis, we capture users' intent through query expansion. We build on earlier work in the area by adopting a pseudo-relevance feedback approach; however, we advance the state of the art by proposing an approach for feature learning within the process of query expansion. In our work, we specifically consider the Wikipedia corpus as the feedback collection space and identify the best features within this context for term selection in two supervised and unsupervised models. We compare our work with state of the art query expansion techniques, the results of which show promising robustness and improved precision.