Project : axis
Section: Scientific Foundations
Keywords : usage mining, web usage mining, data warehouse, data mining, sequential patterns, clustering, user behaviour.
Usage Mining : Applying KDD to Usage Data
Let us consider the KDD process represented by Fig. 2. This process is made of four main steps:
Data selection aims at extraction from the database or datawarehouse the information needed by the data mining step.
Data transformation will then use parsers in order to create data tables which can be used by the data mining algorithms.
Data mining techniques range from sequential patterns to association rules or cluster discovery.
finally the last step will allow the re-use of the obtained results into a usage analysis process.
The studies conducted over KDD applied to usage data have two goals: improving the usage of the IS and/or enhance the IS by comparing the structure information about the IS with the results of the usage analysis.
Let us zoom on the five following research topics:
Data selection and transformation
The considered KDD methods will rely on the notion of session, represented through a tabular model (items), an association rules model (itemsets) or a graph model. This notion of session enables us to act in the good level during the process of knowledge extraction from log files. Our goal is to build summaries and generate statistics on these summaries. At this level of formalization we can consider rules and graphs, define hierarchical structures on variables, extract sequences and thus build new types of data by using KDD methods.
Actually, as the analysis methods come from various research fields (data analysis, statistics, data mining, A.I., ...), a data transformation from input to output is needed and will be managed by the parsers. The input data will come from databases or from standard formatted file (XML) or a private format.
We insist on the importance of this step in the KDD process.
Extracting association rules
Our preprocessing tools (or generalization operators) given in the previous part were designed to build summaries and also generate statistics on these summaries. At this level of formalization we can consider rules and graphs, define hierarchical structures on variables, extract sequences and thus build new types of data by using methods for extracting frequent itemsets or association rules.
These methods were first presented in 1993 by R. Agrawal, T. Imielinski and A. Swami (researchers in databases at the IBM research center, Almaden). They are available in market software for data mining (IBM's intelligent miner or SAS's enterprise miner).
Our approach will rely on work coming from the field of generalization operators and data aggregation. These summaries can be integrated in a recommendation mechanism for the user help. We propose to adapt frequent itemset research methods or association rules discovery methods to the Web Usage Mining problem. We may get inspired by methods coming from the genomist methods (which present common characteristics with our field). If the goal of the analysis can be written in a decisional framework then the clustering methods will identify usage groups based on the extracted rules.
Discovering sequential patterns
Knowing the user can be based on sequential pattern (which are inter transactions patterns) discovery. Sequential patterns offer a strong correlation with Web Usage Mining (and more generally with usage analyzes problems) purposes. Our goal is to provide extraction methods which are as efficient as possible, and also to improve the relevance of their results. For this purpose, we plan to enhance the sequential pattern extraction methods by taking into account the context where those methods are involved. This can be done:
first of all by analyzing the causes of a sequential pattern extraction failure on large access logs. It is necessary to understand and incorporate the great variety of potential behaviours on a Web site. This variety is mainly due to the large size of the trees representing the Web sites and the very large number of combination of navigations on those sites.
It is also necessary to incorporate all the available information related to the usage. Taking into account several information sources in a single sequential pattern extraction process is a challenge and can lead to numerous opportunities.
Finally, sequential pattern mining methods will have to get adapted to a new and growing domain: data streams. In fact, in numerous practical cases, data cannot be stored more than a specified time (and even not at all). Data mining methods will have to provide solution in order to respect the specific constraints related to this domain (no multiple scan over the data, no blocking actions, etc.).
Clustering approach for reducing the volume of data in data warehouses
Clustering is one of the most popular techhnique in knowledge acquisition and it is applied in various fields including data mining and statistical data analysis. This task organizes a set of individuals into clusters in such a way that individual within a given cluster have a high degree of similarity, while individuals belonging to different clusters have a high degree of dissimilarity.
The definition of 'homogeneous' cluster depends on a particular algorithm: this is indeed a simple structure, which, in the absence of a priori knowledge about the multidimensional shape of the data, may be a reasonable starting point towards the discovery of richer and more complex structures
Clustering methods reduce the volume of data in data warehouses, preserving the possibility to perform needed analysis. The rapid accumulation of large databases of increasing complexity poses a number of new problems that traditional algorithms are not equipped to address. One important feature of modern data collection is the ever increasing size of a typical database: it is not so unusual to work with databases containing from a few thousands to a few millions of individuals and hundreds or thousands of variables. Now, most clustering algorithms of the traditional type are severely limited as to the number of individuals they can comfortably handle.
Cluster analysis may be divided into hierarchical and partitioning methods. Hierarchical methods yields complete hierarchy, i.e., a nested sequence of partitions of the input data. Hierarchical methods can be agglomerative or divisive. Agglomerative methods yields a sequence of nested partitions starting with the trivial clustering in which each individual is in a unique cluster and ending with the trivial clustering in which all individuals are in the same cluster. A divisive method starts with all individuals in a single cluster and performs splitting until a stopping criterion is met. Partitioning methods aim at obtaining a partition of the set of individuals into a fixed number of clusters. These methods identify the partition that optimizes (usually locally) an adequacy criterion.
Reusing usage analysis experiences
This topic aims at re-using previous analysis results into current analysis: in the short run we will work on an incremental approach of the discovery of sequential motives; in the longer run our approach will be based upon case-based reasoning. Nowadays very fast algorithms have been developed which efficiently search for dependences between attributes (research algorithms with association rules), or dependences between behaviours (research algorithms with sequential motives) within large databases.
Unfortunately, even though these algorithms are very efficient, and depending on the size of the database, it can sometimes take up to several days to retrieve relevant and useful information. Furthermore, the variation of parameters provided to the user requires to re-start the algorithms without taking previous results into account. Similarly, when new data is added or suppressed from the base, it is often necessary to re-start the retrieval process to maintain the extracted knowledge.
Considering the size of the handled data, it is essential to propose both an interactive (parameters variation) and incremental (data variation in the base) approach in order to rapidly meet the needs of the end user.
This problematic is currently considered as a research problem open within the framework of Data Mining; and even though a few solutions exist, they are not quite satisfactory because they only provide a partial solution to the problem.
