1980–2001

Semantic and Object-Oriented Data Modeling

Semantic grounding and temporal integrity emerged as core principles to preserve meaning across evolving datasets. Object-oriented and ontology-inspired approaches replaced flat relational schemas with richer interdependencies, constraints, and knowledge structures. Spatiotemporal modeling frameworks and probabilistic graphical representations supported complex systems, uncertainty, and cross-domain interoperability. Generalized Stochastic Petri Nets and Bayes Net tools anchored rigorous simulation and inference throughout these developments.

Period 1980-1986: Semantic grounding, temporal integrity, object-oriented CAD structures, NF2 relations, dataflow, and time-series methods drove early modeling[1][2][3][4][5][6][7][8][9][10][11][12][13][14][15]. Foundational spatial partitioning, unified NF2 relations, semantic design, and robust data practices shaped modern GIS, object-relational databases, and analytics[16][17][18][19][20]. Period 1987-1994: Semantic and object-oriented modeling utilized modularity, fuzzy logic, splines, and deductive reasoning to capture complex relations[21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39]. These foundational regularizations, rough sets, multilevel structures, and fuzzy methods shaped modern ontologies, machine learning, and data integration[40][21][41][42][43]. Period 1995-2001: Object-oriented spatiotemporal schemas, fuzzy relational modeling, probabilistic data association, and XML repositories integrated heterogeneous sources and dynamic phenomena[44][45][46][47][48][49][50][51][52][53][54][55][56][57][58][59][60]. Stochastic Petri nets, joint longitudinal models, Bayesian network toolboxes, and covariance structures shaped simulation, prognostic inference, and multivariate analysis[61][62][63][64].

2002–2015

Unified Multimodal and Spatial Data Modeling

Researchers pursued unified data modeling frameworks across relational, multidimensional, and spatiotemporal domains to achieve cross-source consistency. Mathematical techniques like manifold learning and dimensionality reduction enabled low-dimensional embeddings for complex data spaces. The focus shifted toward data-intensive spatial modeling, leveraging hierarchical Bayesian inference and large-scale information fusion for environmental and urban processes. These developments established scalable architectures, sparse representation models, and standardized infrastructures that supported automated knowledge extraction and real-time analytics.

Period 2002-2008: Researchers integrated relational, fuzzy, and spatiotemporal frameworks while advancing manifold reduction, time-series Poisson models, and scalable data summarization[1][2][3][4][5][6][7][8][9][10][11][12][13][14][15]. Entity-relationship models, Laplacian eigenmaps, data-reshaping packages, and t-SNE techniques subsequently anchored modern database design, dimensionality reduction, and exploratory data analysis[1][10][16][17]. Period 2009-2015: Data-intensive spatial modeling utilized hierarchical Bayesian inference, scalable frameworks, and semantic integration for environmental processes[18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35]. Foundational works on data-intensive discovery, sparse representation, spatial Bayesian frameworks, and longitudinal SEM shaped later interdisciplinary research[18][36][24][37].

2016–2023

Physics-Informed and Data-Driven Modeling

This era integrates data-driven techniques with physics-based constraints to enable uncertainty-aware learning and cross-domain calibration. Bayesian inference and data assimilation unify model calibration and uncertainty quantification across scientific disciplines. Reduced-order representations and data compression support efficient simulations of complex dynamical systems for engineering applications. Furthermore, foundational work on latent-variable approaches and mediation analysis shapes how researchers conceive and validate models.

Period 2016-2023: Bayesian inference, reduced-order representations, and latent variables unite physics and data across systems[1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16]. Datasheets, PLS-SEM, mediation analysis, and mechanics frameworks shaped modern data modeling and reproducibility[17][18][13][19].

Semantic and Object-Oriented Data Modeling era

Stephen W. Raudenbush and 2 others

Stephen W. Raudenbush[1] conducted research while affiliated with Michigan State University[4] and Harvard University[5]. Through the work published in Hierarchical Linear Models: Applications and Data Analysis Methods[9], they advanced probabilistic graphical representations that supported complex systems and data modeling. Anthony S. Bryk[2] performed research while based at the University of Chicago[6]. Their contributions to probabilistic frameworks and inference within evolving datasets were established through Hierarchical Linear Models: Applications and Data Analysis Methods[9]. Nicholas T. Longford[3] carried out investigations at the Educational Testing Service[7] and Princeton University[8]. They contributed to the era's thematic focus on complex systems and uncertainty handling through Hierarchical Linear Models: Applications and Data Analysis Methods[9].

Unified Multimodal and Spatial Data Modeling era

Geoffrey E. Hinton and Laurens van der Maaten

Geoffrey E. Hinton[10], while researching at the University of Toronto[12] and University College London[13], contributed to the era's mathematical techniques for dimensionality reduction. Through the paper Visualizing Data using t-SNE[14], this work enabled low-dimensional embeddings for complex data spaces to achieve cross-source consistency. Laurens van der Maaten[11] likewise advanced these manifold learning frameworks for complex data spaces. In the paper Visualizing Data using t-SNE[14], this contribution provided unified data modeling capabilities by mapping intricate structures into low-dimensional embeddings.

Physics-Informed and Data-Driven Modeling era

José L. Roldán and 2 others

José L. Roldán[15] conducted research while affiliated with Universidad de Sevilla[18]. Through Mediation analysis in partial least squares path modeling[21], this work provided foundational insights into latent-variable approaches and mediation analysis that shape how researchers conceive and validate models. Christian Nitzl[16] carried out investigations while based at Universität der Bundeswehr München[19] and Bundeswehr[20]. Via Mediation analysis in partial least squares path modeling[21], this contribution advanced foundational work on latent-variable approaches and mediation analysis to shape how researchers conceive and validate models. Gabriel Cepeda‐Carrión[17] performed scholarly activities while located at Universidad de Sevilla[18]. With Mediation analysis in partial least squares path modeling[21], the research established essential methodological foundations on latent-variable approaches and mediation analysis shaping how researchers conceive and validate models.