Boris Cule

Mining closed strict episodes

Discovering patterns in a sequence is an important aspect of data mining. One popular choice of such patterns are episodes, patterns in sequential data describing events that often occur in the vicinity of each other. Episodes also enforce in which order the events are allowed to occur. In this work we introduce a technique for discovering closed episodes. Adopting existing approaches for discovering traditional patterns, such as closed itemsets, to episodes is not straightforward. First of all, we cannot define a unique closure based on frequency because an episode may have several closed superepisodes. Moreover, to define a closedness concept for episodes we need a subset relationship between episodes, which is not trivial to define. We approach these problems by introducing strict episodes. We argue that this class is general enough, and at the same time we are able to define a natural subset relationship within it and use it efficiently. In order to mine closed episodes we define an auxiliary closure operator. We show that this closure satisfies the needed properties so that we can use the existing framework for mining closed patterns. Discovering the true closed episodes can be done as a post-processing step. We combine these observations into an efficient mining algorithm and demonstrate empirically its performance in practice.

Mining closed episodes with simultaneous events

Sequential pattern discovery is a well-studied field in data mining. Episodes are sequential patterns describing events that often occur in the vicinity of each other. Episodes can impose restrictions to the order of the events, which makes them a versatile technique for describing complex patterns in the sequence. Most of the research on episodes deals with special cases such as serial, parallel, and injective episodes, while discovering general episodes is understudied. In this paper we extend the definition of an episode in order to be able to represent cases where events often occur simultaneously. We present an efficient and novel miner for discovering frequent and closed general episodes. Such a task presents unique challenges. Firstly, we cannot define closure based on frequency. We solve this by computing a more conservative closure that we use to reduce the search space and discover the closed episodes as a postprocessing step. Secondly, episodes are traditionally presented as directed acyclic graphs. We argue that this representation has drawbacks leading to redundancy in the output. We solve these drawbacks by defining a subset relationship in such a way that allows us to remove the redundant episodes. We demonstrate the efficiency of our algorithm and the need for using closed episodes empirically on synthetic and real-world datasets.

A new constraint for mining sets in sequences

Discovering interesting patterns in event sequences is a popular task in the field of data mining. Most existing methods try to do this based on some measure of cohesion to determine an occurrence of a pattern, and a frequency threshold to determine if the pattern occurs often enough. We introduce a new constraint based on a new interestingness measure combining the cohesion and the frequency of a pattern. For a dataset consisting of a single sequence, the cohesion is measured as the average length of the smallest intervals containing the pattern for each occurrence of its events, and the frequency is measured as the probability of observing an event of that pattern. We present a similar constraint for datasets consisting of multiple sequences. We present algorithms to efficiently identify the thus defined interesting patterns, given a dataset and a user-defined threshold. After applying our method to both synthetic and real-life data, we conclude that it indeed gives intuitive results in a number of applications.

Mining Closed Strict Episodes

Discovering patterns in a sequence is an important aspect of data mining. One popular choice of such patterns are episodes, patterns in sequential data describing events that often occur in the vicinity of each other. Episodes also enforce in which order events are allowed to occur. In this work we introduce a technique for discovering closed episodes. Adopting existing approaches for discovering traditional patterns, such as closed item sets, to episodes is not straightforward. First of all, we cannot define a unique closure based on frequency because an episode may have several closed super episodes. Moreover, to define a closedness concept for episodes we need a subset relationship between episodes, which is not trivial to define. We approach these problems by introducing strict episodes. We argue that this class is general enough, and at the same time we are able to define a natural subset relationship within it and use it efficiently. In order to mine closed episodes we define an auxiliary closure operator. We show that this closure satisfies the needed Galois connection so that we can use the existing framework for mining closed patterns. Discovering the true closed episodes can be done as a post-processing step. We combine these observations into an efficient mining algorithm and demonstrate empirically its performance in practice.

Mining train delays

The Belgian railway network has a high traffic density with Brussels as its gravity center. The star-shape of the network implies heavily loaded bifurcations in which knock-on delays are likely to occur. Knock-on delays should be minimized to improve the total punctuality in the network. Based on experience, the most critical junctions in the traffic flow are known, but others might be hidden. To reveal the hidden patterns of trains passing delays to each other, we study, adapt and apply the state-of-the-art techniques for mining frequent episodes to this specific problem.

Pattern Based Sequence Classification

Sequence classification is an important task in data mining. We address the problem of sequence classification using rules composed of interesting patterns found in a dataset of labelled sequences and accompanying class labels. We measure the interestingness of a pattern in a given class of sequences by combining the cohesion and the support of the pattern. We use the discovered patterns to generate confident classification rules, and present two different ways of building a classifier. The first classifier is based on an improved version of the existing method of classification based on association rules, while the second ranks the rules by first measuring their value specific to the new data object. Experimental results show that our rule based classifiers outperform existing comparable classifiers in terms of accuracy and stability. Additionally, we test a number of pattern feature based models that use different kinds of patterns as features to represent each sequence as a feature vector. We then apply a variety of machine learning algorithms for sequence classification, experimentally demonstrating that the patterns we discover represent the sequences well, and prove effective for the classification task.

Mining association rules in long sequences

Discovering interesting patterns in long sequences, and finding confident association rules within them, is a popular area in data mining. Most existing methods define patterns as interesting if they occur frequently enough in a sufficiently cohesive form. Based on these frequent patterns, association rules are mined in the traditional manner. Recently, a new interestingness measure, combining cohesion and frequency of a pattern, has been proposed, and patterns are deemed interesting if encountering one event from the pattern implies with a high probability that the rest of the pattern can be found nearby. It is quite clear that this probability is not necessarily equally high for all the events making up such a pattern, which is why we propose to introduce the concept of association rules into this problem setting. The confidence of such an association rule tells us how far on average from a particular event, or a set of events, one has to look, in order to find the rest of the pattern. In this paper, we present an efficient algorithm to mine such association rules. After applying our method to both synthetic and real-life data, we conclude that it indeed gives intuitive results in a number of applications.

Mining the entire Protein DataBank for frequent spatially cohesive amino acid patterns

BackgroundThe three-dimensional structure of a protein is an essential aspect of its functionality. Despite the large diversity in protein structures and functionality, it is known that there are common patterns and preferences in the contacts between amino acid residues, or between residues and other biomolecules, such as DNA. The discovery and characterization of these patterns is an important research topic within structural biology as it can give fundamental insight into protein structures and can aid in the prediction of unknown structures.ResultsHere we apply an efficient spatial pattern miner to search for sets of amino acids that occur frequently in close spatial proximity in the protein structures of the Protein DataBank. This allowed us to mine for a new class of amino acid patterns, that we term FreSCOs (Frequent Spatially Cohesive Component sets), which feature synergetic combinations. To demonstrate the relevance of these FreSCOs, they were compared in relation to the thermostability of the protein structure and the interaction preferences of DNA-protein complexes. In both cases, the results matched well with prior investigations using more complex methods on smaller data sets.ConclusionsThe currently characterized protein structures feature a diverse set of frequent amino acid patterns that can be related to the stability of the protein molecular structure and that are independent from protein function or specific conserved domains.

MARBLES: Mining association rules buried in long event sequences

Sequential pattern discovery is a well-studied field in data mining. Episodes are sequential patterns that describe events that often occur in the vicinity of each other. Episodes can impose restrictions on the order of the events, which makes them a versatile technique for describing complex patterns in the sequence. Most of the research on episodes deals with special cases such as serial and parallel episodes, while discovering general episodes is surprisingly understudied. This is particularly true when it comes to discovering association rules between them.

Itemset Based Sequence Classification

Sequence classification is an important task in data mining. We address the problem of sequence classification using rules composed of interesting itemsets found in a dataset of labelled sequences and accompanying class labels. We measure the interestingness of an itemset in a given class of sequences by combining the cohesion and the support of the itemset. We use the discovered itemsets to generate confident classification rules, and present two different ways of building a classifier. The first classifier is based on the CBA (Classification based on associations) method, but we use a new ranking strategy for the generated rules, achieving better results. The second classifier ranks the rules by first measuring their value specific to the new data object. Experimental results show that our classifiers outperform existing comparable classifiers in terms of accuracy and stability, while maintaining a computational advantage over sequential pattern based classification.