Constraint-Based Pattern Discovery

Devising fast and scalable algorithms, able to crunch huge amount of data, was for many years one of the main goals of data mining research. But then we realized that this was not enough. It does not matter how efficient such algorithms can be, the results we obtain are often of limited use in practice. Typically, the knowledge we seek is in a small pool of local patterns hidden within an ocean of irrelevant patterns generated from a sea of data. Therefore, it is the volume of the results itself that creates a second order mining problem for the human expert. This is, typically, the case of association rules and frequent itemset mining (Agrawal & Srikant, 1994), to which, during the last decade a lot of researchers have dedicated their (mainly algorithmic) investigations. The computational problem is that of efficiently mining from a database of transactions, those itemsets which satisfy a user-defined constraint of minimum frequency. Recently the research community has turned its attention to more complex kinds of frequent patterns extracted from more structured data: sequences, trees, and graphs. All these different kinds of pattern have different peculiarities and application fields, but they all share the same computational aspects: a usually very large input, an exponential search space, and a too large solution set. This situation—too many data yielding too many patterns—is harmful for two reasons. First, performance degrades: mining generally becomes inefficient or, often, simply unfeasible. Second, the identification of the fragments of interesting knowledge, blurred within a huge quantity of mostly useless patterns, is difficult. The paradigm of constraintbased pattern mining was introduced as a solution to both these problems. In such paradigm, it is the user who specifies to the system what is interesting for the current application: constraints are a tool to drive the mining process towards potentially interesting patterns, moreover they can be pushed deep inside the mining algorithm in order to fight the exponential search space curse, and to achieve better performance (Srikant et al., 1997; Ng et al. 1998; Han et al., 1999; Grahne et al., 2000).