Tracking the evolution of user interests is a problem instance of concept drift learning. Keeping track of multiple interest categories is a natural phenomenon as well as an interesting tracking problem because interests can emerge and diminish at different time frames. The first part of this dissertation presents a Multiple Three-Descriptor Representation (MTDR) algorithm, a novel algorithm for learning concept drift especially built for tracking the dynamics of multiple target concepts in the information filtering domain. The learning process of the algorithm combines the long-term and short-term interest (concept) models in an attempt to benefit from the strength of both models. The MTDR algorithm improves over existing concept drift learning algorithms in the domain. Being able to track multiple target concepts with a few examples poses an even more important and challenging problem because casual users tend to be reluctant to provide the examples needed, and learning from a few labeled data is generally difficult. The second part presents a computational Framework for Extending Incomplete Labeled Data Stream (FEILDS). The system modularly extends the capability of an existing concept drift learner in dealing with incomplete labeled data stream. It expands the learner's original input stream with relevant unlabeled data; the process generates a new stream with improved learnability. FEILDS employs a concept formation system for organizing its input stream into a concept (cluster) hierarchy. The system uses the concept and cluster hierarchy to identify the instance's concept and unlabeled data relevant to a concept. It also adopts the persistence assumption in temporal reasoning for inferring the relevance of concepts. Empirical evaluation indicates that FEILDS is able to improve the performance of existing learners particularly when learning from a stream with a few labeled data. Lastly, a new concept formation algorithm, one of the key components in the FEILDS architecture, is presented. The main idea is to discover intrinsic hierarchical structures regardless of the class distribution and the shape of the input stream. Experimental evaluation shows that the algorithm is relatively robust to input ordering, consistently producing a hierarchy structure of high quality.

In recent years, music has become increasingly universal due to technological advances. All kinds of music have become more complex and a large amount around us. How recommending the music that user is interested in from a wide variety of music is the development intentions of the music recommendation system MRS (Music Recommendation System). In the recommending system, the most widely known is Content-based (CB) and Collaborative (COL). Chen et al. have proposed an alternative way that used CB and COL of music recommendation. The purpose of the CB method is to recommend the music objects that belong to the music groups the user is recently interested in. Each transaction is assigned a different weight, where the latest transaction has the highest weight. The preferences of users are derived from the access histories and recorded in profiles. Based on the collaborative approach, the purpose of the COL method is to provide unexpected findings due to the information sharing between relevant users. But in the CB method, the formula of computing music group weight pays much attention to the weight of the transaction. This will lead to the result that the group weight of music group B which appears once in the later transaction is larger than the group weight of the music group A which appears many times in the earlier transaction. In the COL method, they do not care the density of the group, where high density means that the transactions which the music group appears are close in the access history of the user. This will lead to the result that the supports of the groups which have different densities are the same, and then the users may be grouped together. Therefore, in this thesis, we propose the TICI (Transaction-Interest-Count-Interest) method to improve the CB method. Considering the two situations of the music group that user is interested in, the large count of music group and the appearance in the later transaction, we put two parameters: Count-Interest and Transaction-Interest in our TICI method to let users choose which weight they want to emphasize. Sometimes, people not only want the music object from one group. We extend the TICI method to find the group pair that the user is interested in. We use two thresholds: CountT and WeightT to decide which candidates can be in the large itemset. In our propose method, we have two possible ways to find the result. And we propose the DI (Density-Interest) method to improve the COL method. Our DI method calculates the supports of music groups and consider the distributions of appearances of the music group. From our simulation results, we show that our TICI method could provide better performance than the CB method. Moreover, our DI method also could provide better performance than the COL method.