Líneas de investigación - Benchmark for Statistical User Models

**R-Tree based User Model**

This proposal involves a statistical user model under the collaborative approach. This approach differ from classical user models since it avoid the loss of knowledge of most classical user models. The user model is based on an R-Tree structure and hierarchically divides the space of users into several overlapped sub-spaces.

According to the R-Tree performance, new elements (individual descriptions of user sessions) will be added to the leaves of the tree. The rest of the nodes in the tree (non-leaf nodes) will keep the user.s descriptions representing the set of users in its sub-tree. Therefore, it can be stated that the universe of users in this approach is represented with different detail at each level of the tree, from the root node to the conjunction of all user.s descriptions in the leaves of the tree. Therefore, the most detailed description is found in the conjunction of all users of level n where n is the level of the leaves in the tree.

**Evaluation and Benchmark**

Evaluating a user model is a non-standardized and time consumption task. In fact it is one of the hardest barriers for user model developers. In fact, empirical evaluations are needed to determine how much reliable is the statistical inference on the models built.

A complete framework to evaluate and measure the performance of any user model for user modeling is needed. Thus, we have developed a complete framework to evaluate statistical user models, and a Benchmark to support that evaluation which will be replicable and comparable. In addition, this Benchmark aims to fill the lack of frameworks to evaluate statistical user models for user modeling.