Log Pre-Processing and Grammatical Inference for Web Usage Mining
In: UM 2005 : "Workshop on Machine Learning for User Modeling: Challenges", 24-25 Jul 2005, Edinburgh, Scotland.
In this paper, we propose a Web Usage Mining pre-processing method to retrieve missing data from the server log files. Moreover, we propose two levels of evaluation: directly on reconstructed data, but also after a machine learning step by evaluating inferred grammatical models. We conducted some experiments and we showed that our algorithm improves the quality of user data.