Evaluating the impact of the user profile dimension on its characterization effectiveness: method based on the evaluation of user community organizations quality

This paper studies the effect of the users' profile size (weighted set of keywords) on the efficiency of the measurement. The goal is not directly to identify the better dimension but to understand its influence on the quality of the result. We show that contrary to what common sense suggests, high dimension profile (high level of knowledge about users) reduces the precision of the measurement. To obtain this result, we develop a metric comparison method based on the evaluation of cluster organization quality. One of the main practical applications of this result is that: more than better precision, reduced profiles speed up computation time and reduce needed resources. We also discuss more theoretical possible conclusions and application of our work in the context of community characterization.