Название: Periodic Pattern Mining: Theory, Algorithms, and Applications
Автор: R. Uday Kiran, Philippe Fournier-Viger
Издательство: Springer
Год: 2021
Страниц: 263
Язык: английский
Формат: pdf (true), epub
Размер: 38.7 MB

This book provides an introduction to the field of periodic pattern mining, reviews state-of-the-art techniques, discusses recent advances, and reviews open-source software. Periodic pattern mining is a popular and emerging research area in the field of data mining. It involves discovering all regularly occurring patterns in temporal databases. One of the major applications of periodic pattern mining is the analysis of customer transaction databases to discover sets of items that have been regularly purchased by customers. Discovering such patterns has several implications for understanding the behavior of customers. Since the first work on periodic pattern mining, numerous studies have been published and great advances have been made in this field. The book consists of three main parts: introduction, algorithms, and applications.

Technological advances in the field of Information and Communication Technologies (ICTs) have facilitated organizations to collect, store, and process massive amounts of data. Useful information that can empower the end-users to achieve socio-economic development lies in this data. However, finding interesting information in data can be very challenging due to the sheer scale of the data. In the last decades, researchers from the field of data mining have aimed at tackling this challenge by proposing various techniques to discover knowledge hidden in voluminous real-world data. Over the years, data mining has received more and more attention from both industry and academia. Pattern mining is one of the fundamental knowledge discovery techniques used in data mining. It involves discovering all user interest-based patterns in a database. Much of the past research on pattern mining has focused on utilizing the frequency-based measures to discover different types of interesting patterns such as frequent patterns, correlated patterns, top-k-frequent patterns, maximal frequent patterns, closed frequent patterns, rare patterns, coverage patterns, high utility patterns, and emerging patterns.

Although discovering frequent patterns in a database is beneficial for many applications, frequency may not always be enough to find user interest-based patterns, especially if the data contains temporal information. For example, the user may consider an irregularly occurring frequent pattern to be less interesting over a regularly occurring infrequency (or rare) pattern in the data. Based on this observation, efforts have been put forth in the literature to discover periodically occurring patterns (or periodic patterns) in a temporal database. Since real-world data often contain temporal information, finding periodic patterns has received a great deal of attention. Furthermore, periodic pattern mining has been extended to consider other forms of data, such as quantitative temporal databases and sequences.

The first chapter is an introduction to pattern mining and periodic pattern mining. The concepts of periodicity, periodic support, search space exploration techniques, and pruning strategies are discussed. The main types of algorithms are also presented such as periodic-frequent pattern growth, partial periodic pattern-growth, and periodic high-utility itemset mining algorithm. Challenges and research opportunities are reviewed.

The chapters that follow present state-of-the-art techniques for discovering periodic patterns in (1) transactional databases, (2) temporal databases, (3) quantitative temporal databases, and (4) big data. Then, the theory on concise representations of periodic patterns is presented, as well as hiding sensitive information using privacy-preserving data mining techniques.

The book concludes with several applications of periodic pattern mining, including applications in air pollution data analytics, accident data analytics, and traffic congestion analytics.

Contents:
Introduction to Data Mining
Discovering Frequent Patterns in Very Large Transactional Databases
Discovering Full Periodic Patterns in Temporal Databases
Discovering Fuzzy Periodic Patterns in Quantitative Temporal Databases
Discovering Partial Periodic Patterns in Temporal Databases
Finding Periodic Patterns in Multiple Sequences
Discovering Self-reliant Periodic Frequent Patterns
Discovering Periodic High Utility Itemsets in a Discrete Sequence
Mining Periodic High-Utility Sequential Patterns with Negative Unit Profits
Hiding Periodic High-Utility Sequential Patterns
NetHAPP: High Average Utility Periodic Gapped Sequential Pattern Mining
Privacy Preservation of Periodic Frequent Patterns Using Sensitive Inverse Frequency
Real-World Applications of Periodic Patterns
Insights for Urban Road Safety: A New Fusion-3DCNN-PFP Model to Anticipate Future Congestion from Urban Sensing Data