Oct 27 – 30, 2024
Achat Hotel Karlsruhe City
Europe/Berlin timezone

Exploiting Social Networking Insights for MOF Research: The Black Hole Dataset as a Catalyst for ‎Enhanced ML Applications in Material Science

Oct 28, 2024, 8:00 PM
1h
Karoline, ground floor (Achat Hotel)

Karoline, ground floor

Achat Hotel

Poster Machine learning (ML) applications using existing data repositories Poster Session

Speaker

Mehrdad Jalali

Description

Within the expansive domain of Metal-Organic Frameworks (MOFs), navigating the vast datasets ‎for impactful research has posed significant challenges. Addressing this, our study introduces a ‎groundbreaking methodology through MOFGalaxyNet, employing Social Network Analysis (SNA) to ‎illuminate the structure and dynamics of MOF interactions. The core of our strategy, the Black Hole ‎approach, identifies the most influential MOFs—akin to celestial black holes for their significant pull ‎on surrounding entities. This leads to creating the Black Hole dataset, a curated collection of MOFs ‎identified for their pivotal roles within the network. Through sophisticated SNA, we extract the ‎Black Hole dataset, a concise yet comprehensive assembly of influential MOFs poised for significant ‎breakthroughs in research. The Black Hole dataset, derived from advanced community detection ‎and centrality analysis, also provides a focused, high-value resource for ML applications in MOF ‎research. Utilizing the Girvan-Newman algorithm, we segment MOFGalaxyNet into communities, ‎employing Degree and Betweenness centrality measures to highlight key MOFs. The resultant Black ‎Hole dataset not only streamlines research focus towards MOFs with the highest potential impact ‎but also embodies the FAIR (Findable, Accessible, Interoperable, Reusable) principles, offering a ‎robust foundation for ML-driven advancements in materials science. Applying the Girvan-Newman ‎algorithm for community detection, alongside Degree and Betweenness centrality analyses, ‎facilitates identifying and categorizing Black Hole MOFs within MOFGalaxyNet. This methodology ‎empowers ML in MOF Research by providing ML practitioners in the MOF community with a data-rich, targeted, and technically vetted resource for predictive modeling and algorithm training. ‎

Primary authors

Christof Wöll (Karlsruhe Institute of Technology) Mehrdad Jalali

Presentation materials

There are no materials yet.