Editorial: Recent Advances in Data Mining: Methods, Trends, and Emerging Applications

1. Introduction

This special issue presents a notably broad view of contemporary data mining and its applications. The published work covers a wide range of topics, including online misinformation, geoscience prediction, lyrical text mining, sparse keyword analysis, antiviral peptide discovery, cybersecurity, science mapping and STEM education analytics, process mining in large-scale assessments, and multimodal clinical communication analysis. Together, these contributions show how data mining is no longer confined to a narrow set of methods or applications. Instead, they reflect the growing reach of the field into complex, diverse, and high-impact real-world domains where data-driven insights can support critical decisions. Across the collection, recurring concerns include predictive performance, interpretability, sparse and heterogeneous data, multimodal analysis, and the practical translation of analytics into domain-specific applications.

Data mining has evolved from a limited technical endeavor to a deeply interdisciplinary field. The papers in this special issue show that today’s progress is driven not only by algorithmic novelty, but also by the ability to work effectively with sparse, heterogeneous, multimodal, sequential, and domain-specific data. Whether the task is detecting emotion, estimating subsurface pressure, characterizing antiviral peptide landscapes, or understanding student response behavior, the central challenge now is to turn complex data into insights that are robust and actionable. Despite the diversity of applications, several shared methodological trends can be observed. First is the transition to architecture that combine multiple views on the same problem, for example, multi-stage social media analysis, ensemble learning, cross modal attention, or joint use of graph and metadata networks. The second is visible concern for reliability and interpretability in the form of Bayesian confidence estimation, rule based explaninability and structured modeling pipelines. The third characteristic of this collection is its clear focus on practical relevance and real-world impact. What makes these contributions especially compelling is that they are not simply benchmark-driven experiments, but investigations shaped by real and often complex environments, ranging from social media and geoscience to healthcare, education, and bioinformatics. The Special Issue was designed to explore “recent advances in data mining” from both methodological and application perspectives, inviting work on scalable algorithms, semantic and text analysis, graph mining, interpretability, and practical deployment. The published contributions clearly align with this evolving vision.

2. An Overview of Published Articles

To better understand the range of contributions in this special issue, the published work can be viewed as four broad themes: communication and text centered mining, predictive and explainable analytics in high stakes domains, structure aware discovery through networks and sequences, and learning analytics and science mapping, as shown in Table 1. Although each study explores its own unique problem, this way of organizing them help bring up common ideas, practical challenges and emerging future directions that collectively shows how the field of data mining continues to evolve.

Table 1. An overview of contributions of the special issue.

	Title	Authors	Summary	Theme Cluster
1	MSDSI-FND: Multi-Stage Detection Model of Influential Users’ Fake News in Online Social Networks	Hala Al-Mutair and Jawad Berri	Proposes a staged misinformation-detection framework for influential X users using linguistic, user-profile, and propagation features on 68 root tweets and 40,000+ engagements.	Communication and text-centered mining
2	Emotion in Words: The Role of Ed Sheeran and Sia’s Lyrics on the Musical Experience	Catarina Travanca, Mónica Cruz and Abílio Oliveira	Uses text mining to compare emotional and thematic patterns in the lyrics of Ed Sheeran and Sia.	Communication and text-centered mining
3	Sparse Keyword Data Analysis Using Bayesian Pattern Mining	Sunghae Jun	Develops Bayesian Pattern Mining to derive more stable keyword associations from highly sparse patent-keyword matrices with credible intervals for lift and confidence.	Communication and text-centered mining
4	From Patient Emotion Recognition to Provider Understanding: A Multimodal Data Mining Framework for Emotion-Aware Clinical Counseling Systems	Saahithi Mallarapu, Xinyan Liu, Pegah Zargarian, Seyyedeh Fatemeh Mottaghian, Ramyashree Suresha, Vasudha Jain and Akram Bayat	Presents a bidirectional framework linking patient emotion recognition and provider behavior analysis using ClinicalBERT, DeBERTa, WavLM, cross-modal attention, and counseling datasets.	Communication and text-centered mining
5	Machine Learning Models for Subsurface Pressure Prediction: A Data Mining Approach	Muhammad Raiees Amjad, Rohan Benjamin Varghese and Tehmina Amjad	Introduces a hybrid meta-ensemble over CNN, RNN, DFNN, RF, and XGBoost for pore-pressure prediction from well logs, reaching an $R^2=0.93$.	Predictive and explainable analytics
6	Rule-Based eXplainable Autoencoder for DNS Tunneling Detection	Giacomo De Bernardi, Giovanni Battista Gaggero, Fabio Patrone, Sandro Zappatore, Mario Marchese and Maurizio Mongelli	Demonstrates an explainable, rule-based alternative to black-box autoencoder approaches for DNS tunneling detection with comparable performance.	Predictive and explainable analytics
7	Mapping the Chemical Space of Antiviral Peptides with Half-Space Proximal and Metadata Networks Through Interactive Data Mining	Daniela de Llano García, Yovani Marrero-Ponce, Guillermin Agüero-Chapin, Hortensia Rodríguez, Francesc J. Ferri, Edgar A. Márquez, José R. Mora, Felix Martinez-Rios and Yunierkis Pérez-Castillo	Uses threshold-free HSPNs and metadata networks in StarPep to map AVP chemical space and identify representative scaffolds and motifs.	Structure-aware discovery
8	Exploring Slow Responses in International Large-Scale Assessments Using Sequential Process Analysis	Daniel Jerez, Elisabetta Mazzullo and Okan Bulut	Examines slow responders in ILSAs using sequence analysis and clustering of PISA 2012 log-file data to identify finer-grained behavioral subgroups.	Structure-aware discovery
9	Quantum Computing in Data Science and STEM Education: Mapping Academic Trends and Analyzing Practical Tools	Eloy López-Meneses, Jesús Cáceres-Tello, José Javier Galán-Hernández and Luis López-Catalán	Combines science mapping of 281 Scopus-indexed publications with tool analysis of platforms such as Qiskit, Quantum Inspire, QuTiP, and Amazon Braket.	Learning analytics and science mapping

2.1. Communication and Text Centered Mining

Contribution 1 by Al-Mutair and Berri address the fake news diffusion by specifically focusing on influential users. They integrate content, user profile and propagation cues in a staged framework, showing that hierarchical feature integration improves detection over simple baseline. Travanca, Cruz and Oliveira in contribution 2 turn text mining toward contemporary music, using lyrical analysis to reveal distinct emotional and thematic patterns in the lyrics of Ed Sheeran and Sia. Jun’s Bayesian pattern mining work in contribution 3 advances sparse keyword analysis by embedding association rule-mining in a probabilistic framework that reports expected interestingness and credible intervals. Contribution 4 by Mallarapu and colleagues extend communication-centered data mining into the clinical domain through a bidirectional framework that couples patient emotion recognition with provider behavior analysis. They applied ClinicalBERT for multi-label patient-side classification and DeBERTa/WavLM with cross-modal attention for provider-side multimodal analysis.

2.2. Predictive and Explainable Analytics

Contribution 5 by Amjad and colleagues show how ensemble-based data mining can improve pore-pressure prediction from well-log data in the Potwar Basin of Pakistan. Their hybrid-meta-ensemble combines deep learning and tree-based learning methods. They reported strong predictive performance suggesting that ensemble learning can improve predictive robustness over single models. De Bernardi and co-authors focus on interpretable anomaly detection in contribution 6. Their proposed rule-based explainable autoencoder for DNS tunnelling detection seeks performance comparable to black-box approaches. They provide analysts clearer insights into system decisions and facilitate human inspection and rule refinement when required. These studies reflect an important shift in applied data mining, where achieving high accuracy is increasingly expected to go hand in hand with transparency and meaningful use within specific domains.

2.3. Structure Aware Discovery

Contribution 7 by De Llano Garcia and co-authors map antiviral peptide chemical space using half space proximal networks and metadata networks. They identified chemically and biologically distinct communities, compact representative scaffold sets, and a set of candidate motifs that includes both previously reported and apparently novel patterns. Jerez and colleagues in contribution 8 use sequence analysis and clustering of PISA 2012 log-file data in the context of international large-scale educational assessments. They showed that “slow responders” should not be treated as a single homogeneous group. Instead, the action sequences can reveal more nuanced behavioral subtypes. Despite their differences in domain, both studies point to the same methodological lesson, i.e., meaningful patterns are often better captured when the inherent structure of the data is preserved, rather than compressed too quickly into summary features.

2.4. Learning Analytics and Science Mapping

Contribution 9 by Lopez-Meneses and co-authors contribute a science-mapping and tool-analysis study focused on quantum computing in data science and STEM education. They examined 281 Scopus indexed publications and reviewed major educational platforms including Qiskit, Quantum Inspire, QuTiP, and Amazon Braket. Their work positions data mining not only as a modeling toolkit but also as a meta-research instrument for understanding how new technical fields emerge, cluster, and translate into educational practice.

3. Emerging Research Direction

Several emerging future directions and challenges are reported by the contributions of this special issue. First challenge is the cross-domain validation. Papers have reported strong results on tightly defined datasets, but generalization across datasets, environments, and institutions remain an open challenge, whether in geoscience, cybersecurity, counseling, or educational assessments. Second challenge is uncertainty aware learning under difficult data conditions, especially for sparse, imbalanced, or long-tailed problems. The Bayesian and network-based approaches in this issue suggest that reliability estimation should become a standard design feature rather than an afterthought. A third emerging direction in interpretable multimodal analytics, where focus is not only on combining different data types but also on understanding how text, audio, structure work together to lead to decisions. A fourth area is longitudinal and sequential modeling because many real-world applications involve the evolving interactions over time rather than isolated snapshots. A fifth concern that becomes increasingly important in several application areas represented in this Special Issue is ethical deployment, specifically in social, educational, and clinical settings. These are the areas where usefulness of data mining methods relates to issues of privacy, fairness, and responsible human oversight. Finally, a sixth direction points towards human in the loop discovery. The domain experts play an active role in guiding scaffold exploration, refining interpretable rules, and contextualizing discovered patterns. Collectively, these directions suggest how the field of data mining is continuing to evolve based on the patterns that emerge across the studies in this collection.