A Practical Algorithm to Find the Best Episode Patterns

Hirao, Masahiro; Inenaga, Shunsuke; Shinohara, Ayumi; Takeda, Masayuki; Arikawa, Setsuo

doi:10.1007/3-540-45650-3_37

Cited by 21 publications

(22 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We declare that this work generalizes and outperforms the ones accomplished in [8,9], since it is capable of discovering more advanced and useful patterns. In fact, we show some experimental results that convince us of the accuracy of our algorithms as well as their fast performances.…”

Section: Introductionmentioning

confidence: 73%

“…Against the fact that finding the best subsequence patterns to separate given two sets of strings is NP-hard, they proposed an algorithm to solve the problem with practically reasonable performance. More recently, an efficient algorithm to discover the best episode patterns was proposed in [9]. An episode pattern p, k , where p is a string and k is an integer, is said to match a string w if p is a subsequence of a substring u of w with |u| ≤ k [14,6,20].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Discovering Best Variable-Length-Don’t-Care Patterns

et al. 2002

Self Cite

View full text Add to dashboard Cite

Abstract. A variable-length-don't-care pattern (VLDC pattern)is an element of set Π = (Σ ∪ { }) * , where Σ is an alphabet and is a wildcard matching any string in Σ * . Given two sets of strings, we consider the problem of finding the VLDC pattern that is the most common to one, and the least common to the other. We present a practical algorithm to find such best VLDC patterns exactly, powerfully sped up by pruning heuristics. We introduce two versions of our algorithm: one employs a pattern matching machine (PMM) whereas the other does an index structure called the Wildcard Directed Acyclic Word Graph (WDAWG). In addition, we consider a more generalized problem of finding the best pair q, k , where k is the window size that specifies the length of an occurrence of the VLDC pattern q matching a string w. We present three algorithms solving this problem with pruning heuristics, using the dynamic programming (DP), PMMs and WDAWGs, respectively. Although the two problems are NP-hard, we experimentally show that our algorithms run remarkably fast.

show abstract

Section: Introductionmentioning

confidence: 73%

Section: Introductionmentioning

confidence: 99%

Discovering Best Variable-Length-Don’t-Care Patterns

et al. 2002

Self Cite

View full text Add to dashboard Cite

show abstract

“…The basic idea of the enumeration is similar to previous works [2][3][4][5][6]8]. The main contrivance of this paper is in the method for calculating the lower bound of the RSS score for specific subspaces of the pattern space.…”

Section: Branch and Bound Strategymentioning

confidence: 91%

“…Another situation is when we are given two sets of strings, where one set (positive set) consists of sequences known to possess some biological characteristic, while the other (negative set) consists of sequences known not to posses these characteristics. The problem is to find a discriminating pattern, that is, a pattern which occurs in most strings of the positive set, but does not occur in most of the strings of the negative set [2][3][4][5][6].…”

Section: Introductionmentioning

confidence: 99%

Practical Algorithms for Pattern Based Linear Regression

et al. 2005

Self Cite

View full text Add to dashboard Cite

Abstract. We consider the problem of discovering the optimal pattern from a set of strings and associated numeric attribute values. The goodness of a pattern is measured by the correlation between the number of occurrences of the pattern in each string, and the numeric attribute value assigned to the string. We present two algorithms based on suffix trees, that can find the optimal substring pattern in O(N n) and O(N 2 ) time, respectively, where n is the number of strings and N is their total length. We further present a general branch and bound strategy that can be used when considering more complex pattern classes. We also show that combining the O(N 2 ) algorithm and the branch and bound heuristic increases the efficiency of the algorithm considerably.

show abstract

“…In particular, these unbounded episodes can be very useful in contexts such as the classification of documents or the intrusion detection systems. As argued in [5], a drawback of subsequence patterns is that they are not suitable for classifying long strings over small alphabet, since a short subsequence pattern matches with almost all long strings. So, the larger the episodes found in a text the better for the future predictions.…”

Section: Advantages Of Our Approachmentioning

confidence: 99%

Discovering Unbounded Episodes in Sequential Data

Casas-Garriga

2003

Knowledge Discovery in Databases: PKDD 2003

View full text Add to dashboard Cite

Abstract. One basic goal in the analysis of time-series data is to find frequent interesting episodes, i.e, collections of events occurring frequently together in the input sequence. Most widely-known work decide the interestingness of an episode from a fixed user-specified window width or interval, that bounds the length of the subsequent sequential association rules. We present in this paper, a more intuitive definition that allows, in turn, interesting episodes to grow during the mining without any user-specified help. A convenient algorithm to efficiently discover the proposed unbounded episodes is also implemented. Experimental results confirm that our approach results useful and advantageous.

show abstract

A Practical Algorithm to Find the Best Episode Patterns

Cited by 21 publications

References 6 publications

Discovering Best Variable-Length-Don’t-Care Patterns

Discovering Best Variable-Length-Don’t-Care Patterns

Practical Algorithms for Pattern Based Linear Regression

Discovering Unbounded Episodes in Sequential Data

Contact Info

Product

Resources

About