Introduction
This special issue of the journal Data Science and Engineering (DSE) presents a curated selection of extended research from the 23rd International Conference on Database Systems for Advanced Applications (DASFAA 2018). The collection reflects the ongoing evolution of the database field into the broader domain of Data Science, showcasing six papers that push the boundaries of large-scale graph computation, secure cloud querying, AI-driven recommendations, and longitudinal healthcare analytics.
The Shift to Advanced Applications
Traditional database research focused primarily on storage and retrieval. However, as highlighted by guest editors Shazia Sadiq and Jianxin Li, the community is now tackling the complexities of "Advanced Applications." This involves managing non-traditional data structures, such as "dual networks" with multi-typed relationships, and ensuring security in outsourced cloud environments. The core challenge is building systems that are simultaneously scalable, secure, and intelligent.
Diverse Methodologies in Modern Data Engineering
The featured papers employ a variety of sophisticated techniques to address these challenges:
- Graph Theory & Algorithms: Development of $k$-connected core models to identify robust structures within massive, multi-layered networks.
- Discrete Hashing: Compression of high-level visual features into compact codes, enabling rapid fashion recommendations in e-commerce contexts.
- Cryptographic Engineering: A novel encryption scheme for SQL range predicates that allows querying of encrypted data while remaining resilient against server collusion.
- Ontological Pattern Matching: A unified fact-checking model that combines knowledge graph triple patterns with "ontology closeness" to verify information accuracy.
- Evolutionary Active Clustering: An interactive framework for medical data that uses "temporal smoothness" constraints to track the evolution of patient clusters (e.g., for Sleep Apnea) over long periods.
Key Insights and Technical Advancements
A critical requirement for this special issue was the "30% Rule," where authors were required to add significant new material and more rigorous evaluation to their original conference papers to ensure deeper technical depth.
Key insights from the collection include:
- Collusion Resistance: In cloud security, it is no longer sufficient to simply encrypt data; systems must be designed to resist attacks where multiple parties share information to compromise privacy.
- Human-in-the-Loop Analytics: For sensitive domains like healthcare, purely automated algorithms are often insufficient. Active learning allows experts to refine how models interpret complex temporal changes in patient data.
- Approximate Query Processing (AQP): As data scales toward the petabyte level, the community is shifting toward AQP, where fast, statistically guaranteed "mostly-correct" answers are prioritized over high-latency exact results.
Results and Impact
The selection process was highly rigorous. From 360 original submissions to DASFAA 2018, only 83 were accepted as full papers (approximately a 23% acceptance rate). From that group, only five were invited to provide extended versions for this issue, alongside one comprehensive survey. These papers demonstrate state-of-the-art performance in computational efficiency for graph analysis and recommendation systems, as well as improved security for range queries compared to existing methods.
Future Directions
The editorial points toward AQP as a major frontier for handling massive datasets. Additionally, maintaining "temporal smoothness" in data mining remains a challenge—ensuring that models tracking data over time do not lose coherence due to noise. The work presented here lays the groundwork for more robust, intelligent, and secure data engineering systems capable of meeting the demands of modern advanced applications.
Summary updated