A Survey of Utility-Oriented Pattern Mining (1805.10511v2)

Abstract: The main purpose of data mining and analytics is to find novel, potentially useful patterns that can be utilized in real-world applications to derive beneficial knowledge. For identifying and evaluating the usefulness of different kinds of patterns, many techniques and constraints have been proposed, such as support, confidence, sequence order, and utility parameters (e.g., weight, price, profit, quantity, satisfaction, etc.). In recent years, there has been an increasing demand for utility-oriented pattern mining (UPM, or called utility mining). UPM is a vital task, with numerous high-impact applications, including cross-marketing, e-commerce, finance, medical, and biomedical applications. This survey aims to provide a general, comprehensive, and structured overview of the state-of-the-art methods of UPM. First, we introduce an in-depth understanding of UPM, including concepts, examples, and comparisons with related concepts. A taxonomy of the most common and state-of-the-art approaches for mining different kinds of high-utility patterns is presented in detail, including Apriori-based, tree-based, projection-based, vertical-/horizontal-data-format-based, and other hybrid approaches. A comprehensive review of advanced topics of existing high-utility pattern mining techniques is offered, with a discussion of their pros and cons. Finally, we present several well-known open-source software packages for UPM. We conclude our survey with a discussion on open and practical challenges in this field.

• The paper presents a structured taxonomy of utility-oriented pattern mining approaches, covering Apriori-based, tree-based, projection, and utility list methods.
• It details how tree-based and projection approaches enhance memory efficiency and reduce computational costs compared to traditional methods.
• It identifies future challenges, including dynamic environments and distributed algorithm designs, to advance scalable, real-time data mining.

A Survey of Utility-Oriented Pattern Mining

The paper "A Survey of Utility-Oriented Pattern Mining" authored by Wensheng Gan et al., provides a comprehensive overview of recent advancements in the field of utility-oriented pattern mining (UPM). This survey categorizes and reviews various state-of-the-art approaches, techniques, and methodologies in UPM, offering insights into the evolution of this important area of data mining. Its aim is to present a structured overview that is useful for researchers working with utility-based data mining applications.

Utility-oriented pattern mining, an extension of traditional pattern mining methods, introduces the notion of utility to address the practical application's demand for discovering patterns that are not only frequent but also meaningful in terms of profit or importance. This concept has significantly expanded pattern mining applications in areas such as cross-marketing, e-commerce, and many other domains requiring value-oriented pattern extraction.

Key Contributions and Taxonomy

The paper provides a structured taxonomy of existing approaches to utility-oriented mining. These approaches are broadly categorized into Apriori-based approaches, tree-based pattern-growth approaches, projection-based approaches, and data-format-based approaches. This categorization reflects innovations to overcome challenges such as candidate generation inefficiencies and the high computational costs associated with early algorithmic developments.

1. Apriori-Based Approaches: Algorithms like UMining, Two-Phase, and their derivatives define the early stage of UPM, where the goal was to extend the principle of frequent pattern mining to utility patterns using candidate generation-and-test methods derived from the Apriori algorithm.
2. Tree-Based Pattern-Growth Approaches: The development of models such as UP-Growth and IHUP shifted the paradigm towards tree-based storage structures, enhancing memory efficiency and reducing computational overhead by leveraging compact data representation.
3. Projection-Based Approaches: These approaches, including CTU-PRO, emphasize the reduction of computational space by storing projected databases and focusing computations only on smaller subsets, leading to improved scalability, particularly for larger datasets.
4. New Data Format-Based Approaches: Techniques employing utility lists (e.g., HUI-Miner, FHM) offer more flexible data structures, efficiently handling high-dimensional data. These approaches benefit from the reduced need for candidate generation and one-phase mining structures.

Advanced Topics and Future Challenges

Despite substantial progress, the paper identifies several key challenges and avenues for future research. Advanced topics examined include the development of utility-oriented methods for dynamic environments, privacy-preserving utility mining, and real-time pattern mining, reflecting the need for more adaptive models capable of operating over data streams and large-scale datasets.

The survey outlines challenges in dealing with domain-specific applications and seeking universal frameworks that can incorporate various forms of patterns. Addressing these challenges involves extending modeling frameworks to integrate domain knowledge effectively and ensuring scalability, especially in the era of Big Data. There is a focus on designing parallel and distributed algorithms effectuated across infrastructures such as MapReduce and Spark, aiming to alleviate performance bottlenecks posed by large and high-velocity data sources.

Conclusion

Utility-oriented pattern mining presents a significant shift from frequency-centric to value-centric pattern extraction methods. Through this survey, researchers are provided with a rich landscape of methodological advancements and challenges that shape the future trajectory of this area. Moving forward, overcoming the intricacies involved in dealing with large and complex datasets while maintaining the relevancy of discovered patterns in practical applications remains a central focus for the data mining community. The potential of UPM to transform industries through actionable insights underscores its significance and the ongoing need for innovative solutions.