Machine Learning-Driven Design and Optimization of Neodymium-Substituted R-Type Hexagonal Ferrites for Enhanced Electrical Polarization Properties

399 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) Machine Learning-Driven Design and Optimization of Neodymium-Substituted R-Type Hexagonal Ferrites for Enhanced Electrical Polarization Properties https://currentsign journal.com/index. php/JCS/index Fahad Rasheed Saba Batool Dr. Engr. Muhammad Anser Kazim Syed Tahir Ali Shah Vol. 3 No. 4 (2025) 400 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) Machine Learning-Driven Design and Optimization of NeodymiumSubstituted R-Type Hexagonal Ferrites for Enhanced Electrical Polarization Properties Multiferroic materials of the R-type hexagonal ferrites have simultaneous magnetic and electrical properties, which is why they are technologically important. Most recently, a promising trend in the control of the structural symmetry and polarization processes within these systems is the replacement of rare-earths by neodymium (Nd3 +). The present-day empirical studies offer the design and optimization of Nd-substituted Ba2Co2Fe16O27-type R-phase ferrites which are formed in a sol-gel auto-combustion process. The experimental data (n=420 samples) comprising of compositional ratios, sintering temperatures, descriptors of grain morphology, and dielectric parameters was trained on an ensemble model with a supervisor applying Random Forest Regression and Gradient-Boosting Algorithms. It was predicted and minimized polarization magnitude (P) and dielectric loss (tan d) with R2 = 0.963 on invisible information which is much better than the conventional regression techniques. At moderate levels of neodymium replacement (x 0.10-0.15), X-ray diffraction (XRD), scanning electron microscopy (SEM), and ferroelectric hysteresis loop analysis validation provided a gain in lattice distortion to allow spontaneous polarization without magnetic order decay. Machine-learning optimization further revealed a nonlinear coupling between macrostrain and domain wall motion, establishing data-driven processing windows for high-performance multiferroic ceramics. Results show that artificial intelligence (AI) can successfully fill in structure-property gaps in complex oxides and minimize the experimental trial-and-error process and allow rational design of state-of-the-art ferrites. The research will offer practical findings to researchers who aim at developing sustainable and smart materials engineering since it can be concluded that algorithmic discovery is supplementary but not substitutive of physical experimentation. Fahad Rasheed * Department of Physics, University of Agriculture, Faisalabad Email: fahadrasheedqadr[email protected] Saba Batool Department of Solid State Physics, Punjab university, Lahore Email: saba.batool1[email protected] Dr. Engr. Muhammad Anser Kazim Department of Mechanical, Electrical & Public Health, Principal Consultant, Meinhardt Group Email: anser[email protected] Syed Tahir Ali Shah TEMA - Centre for Mechanical Technology and Automation, Department of Mechanical Engineering, University of Aveiro, Campus Universitário de Santiago, 3810-193 Aveiro, Portugal Email: [email protected] Abstract 401 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) Keywords: Machine Learning, Neodymium Substitution, Hexagonal Ferrites, Polarization Properties, R-Type Structure, Materials Optimization, Multiferroics Introduction: In the last decade, the fast development of machine learning (ML) methods has revolutionized the materials discovery, modeling, and optimization (Butler et al., 2018; Ramprasad et al., 2017). Conventional experimental procedures of fabricating functional oxides usually focus on compositional modifications and guesswork of the interpretation of structure-property relationships. Such methods are effective but costly and time consuming besides being constrained by human bias. The alternative proposed by ML frameworks is to identify the latent associations among multidimensional datasets that allow predictive modeling and smart experimentation (Ward and Wolverton, 2020). The hexagonal ferrites that combine magnetic and electric orders, and are especially R-type barium ferrites with the general formula Ba2M2Fe16O27 (M = Co2+, Zn2+), are some of the most versatile functional oxide multiferroic materials and are useful to deliver microwave absorption, spintronics, sensors and data-storage devices (Pullar, 2012; Kumar & Singh, 2021).The R-type Ba₂Co₂Fe₁₆O₂₇ structure exhibits ferroelectricity caused by spin-induced polarization associated with noncollinear magnetic ordering (Cheng et al., 2018). However, undoped systems show modest polarization and limited dielectric tunability, restricting their broader utility. To overcome this, rare-earth substitution, particularly with neodymium (Nd³⁺), has been explored to modify local crystal fields and Fe³⁺– O²⁻–Fe³⁺ interactions. Owing to its larger ionic radius (1.109 Å) and unpaired 4f electrons, Nd³⁺ induces lattice strain, affecting magnetic anisotropy and polarization (Sun et al., 2022). Limited Nd substitution enhances dielectric constant and reduces loss tangent through polarization hopping and microstructural refinement (Iqbal et al., 2023), though optimizing doping level, calcination temperature, and homogeneity remains complex. ML algorithms can address this challenge by predicting optimal synthesis parameters and modeling nonlinear relationships between composition, structure, and polarization (Jha et al., 2019). Despite progress in multiferroic synthesis, the rational design of hexagonal ferrites with tailored polarization is limited by the absence of predictive links among composition, structure, and functionality. The doping in the middle range of the Nd-substituted ferrites improves polarization by deforming the Fe-O octahedra but over doping interferes with the stability and magnetic order. The trade-off needs data-driven methods that can quantify effects of competing effects and also, it can determine the optimal conditions. Such relationships may be elicited using regression and feature-importance mapping methods in the framework of ML and provide useful data relating to prediction that cannot be easily identified by a person with his intuition. The optimization of ferrite properties can be accelerated by using ML with experimental validation with the least wastage of material and research time. In this connection, the research question will be as follows: Can ML frameworks predict and optimize the electrical polarization behaviors of neodymium-substituted R-type hexagonal ferrites? The answer to this question will also advance AI-assisted experimentation and contribute to successful discovery of multiferroic materials. The relevance of this work lies in the fact that the materials science, data analytics, and computational modeling have been integrated interdisciplinarily.ML-based materials engineering 402 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) is underdeveloped in Pakistan and the rest of the Global South. The creation of local expertise in materials research with AI assistance can lower costs and enhance the process of reproducibility and thus allow the institutions in the region to engage in intelligent manufacturing (Khan and Siddique, 2024). Optimizing the ferrites technologically to have a high polarization is important in wireless communication, magnetoelectric sensors, and systems that store energy. The given research also shows that small labs can successfully apply open-source ML frameworks, including scikit-learn and TensorFlow, to work with small datasets. It offers an effective way of data-centric innovation within resource constraints because it verifies the predictions by experimentally synthesizing them. Finally, the combination of ML-based prediction and experimental validation is a disruptive method of faster discovering and optimization of multiferroic materials with tunable properties. Objectives and Hypotheses The overall objective of the study is to create a scientifically legitimate, data-informed model of design and optimization of neodymium (Nd)-substituted R-type hexagonal ferrites, and have the objectives of enhancing electrical polarization properties of the same through machine learning algorithms and experimental synthesis.The study focuses on a twofold strategy, which is empirical validation of synthesis and structural characterization, and predictive optimization of parameters. Objectives To prepare neodymium-substituted R-type hexagonal ferrites (Ba2Co2-NdFe16O27) by a controlled sol-gel auto-combustion process and examine their phase purity, morphology and microstructure characteristics. To build machine-learning models based on the algorithms of Random Forest Regression (RFR) and Gradient Boosting (GB) to predict the polarization of the samples and their dielectric properties in relation to compositional and processing descriptors. To determine the statistical dependence of the neodymium concentration, sintering temperature and electrical polarization by use of correlation and regression analysis methods. To identify the optimal neodymium substitution level that yields the highest polarization response with minimum dielectric loss, validated experimentally. To propose a methodological framework combining experimental and computational approaches suitable for resource-constrained laboratories in developing contexts. Research Questions How effectively can machine learning models predict the polarization behavior of Nd-substituted R-type hexagonal ferrites using limited experimental datasets? What is the relationship between compositional variation (Nd substitution) and electrical polarization properties in synthesized ferrites? Does the optimized substitution level correspond with the theoretical prediction of polarization enhancement through lattice distortion? To what extent can computational prediction reduce experimental iteration and cost compared to traditional optimization methods? 403 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) Hypotheses Based on the theoretical and empirical foundation of ferrite materials and machine learning literature, the following hypotheses are proposed: H₁: Neodymium substitution significantly enhances electrical polarization properties of R-type hexagonal ferrites. H₂: Machine-learning models can accurately predict polarization behavior using structural and compositional input variables. H₃: Moderate neodymium substitution (x ≈ 0.10–0.15) optimizes polarization by inducing balanced lattice distortion and micro strain. H₄: Experimental results validate machine-learning predictions with statistically significant correlation (r ≥ 0.80). H₅: The integration of machine learning reduces the experimental optimization cycle time and resource consumption in materials design. Literature Review The literature review section critically analyzes previous studies related to machine-learning applications in materials science, structural and electrical properties of hexagonal ferrites, and the specific influence of rare-earth element substitution. The section synthesizes existing knowledge to identify research gaps addressed by this study. Machine Learning and Materials Science Machine learning has transformed the materials science field to enable quicker discovery, design, and performance forecasting of high-technology materials (Butler et al., 2018; Ramprasad et al., 2017). The old-fashioned trial and error paradigm has slowly been replaced by data-based modeling, the algorithms learn concealed relationships in high-dimensional datasets. Researchers have used supervised learning models like the Random Forests and Neural Networks to predict properties including band gaps, formation energies and dielectric constants (Ward et al., 2016; Jha et al., 2019). Agrawal and Choudhary (2016) define materials informatics as the incorporation of experimental and computational data based on machine-learning algorithm to form property-compositionprocessing relationships. In a complex oxide system, this is especially useful given the nonlinear nature of the interactions between synthesis parameters, crystal orientation and functional behavior. The role of machine learning is also taking a more center stage in the Materials Genome Initiative (MGI), an international initiative that features the reduction of the discovery of new materials by half through time and cost (Kalidindi & De Graef, 2015). By combining openaccess datasets with predictive models, MGI-inspired workflows have successfully optimized semiconductors, superconductors, and thermoelectric compounds. However, ferrites—especially hexagonal phases—remain relatively underrepresented in such frameworks due to their multisublattice complexity and anisotropy. Recent advancements show that even small datasets, when augmented through domain knowledge and proper feature engineering, can yield highly predictive results (Ward & Wolverton, 2020). Therefore, this study positions itself within the emerging domain of machine- 404 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) learning-assisted ferrite optimization, bridging computational and experimental methods in the context of rare-earth substituted materials. Structural and Electrical Characteristics of R-Type Hexagonal Ferrites R-type hexagonal ferrites belong to a family of magnetic oxides with general formula Ba₂M₂Fe₁₆O₂₇, where M is a divalent metal ion such as Co, Zn, or Ni. These compounds are structurally complex, consisting of alternating spinel (S) and hexagonal (R) blocks arranged along the c-axis (Pullar, 2012). Their physical behavior is governed by super exchange interactions among Fe³⁺ ions mediated through oxygen atoms, leading to ferrimagnetism coupled with weak ferroelectricity. The electrical polarization in R-type ferrites arises from asymmetric displacement of Fe³⁺ ions and spin-induced charge redistribution, a mechanism distinct from classical perovskite ferroelectrics (Cheng et al., 2018). The coexistence of electric and magnetic orders renders these materials multiferroic, with potential applications in microwave absorbers, sensors, and magnetoelectric devices. However, the spontaneous polarization of pristine R-type ferrites remains limited (~0.01–0.05 µC/cm²) due to structural rigidity and low domain mobility (Sun et al., 2022). Consequently, researchers have focused on structural modification via cation substitution to enhance ferroelectric and dielectric behavior without deteriorating magnetic order. Among various dopants, rare-earth elements such as Nd, Sm, and La are particularly effective due to their large ionic radii and unpaired 4f electrons, which introduce local lattice distortions and defect dipoles (Iqbal et al., 2023). Influence of Rare-Earth (Neodymium) Substitution Replacement of rare-earth alters the ferrite behavior by having an ionic radius distortion and charge compensation effects. Replacement of Fe3+ by Nd3+ on the lattice parameters, creates micro strain, and affects the grain morphology (Sharma et al., 2021). A moderate level of substitution (x [?] 0.15) enhances polarization through an increase in lattice distortion, spacecharge polarization, and a decrease in the leakage current (Li et al., 2020). However, over substitution may lead to the formation of second phase and replacement of the magnetic paths of exchange which leads to deterioration of the performance. Sattar et al.'s empirical study on Ba₂Co₂₋ₓNdₓFe₁₆O₂₇ ferrites in 2023 showed that polarization peaked at x = 0.12 and then declined as a result of grain boundary scattering. The defectchemistry model, which suggests an ideal substitution range where lattice flexibility encourages polarization without jeopardizing crystal stability, is consistent with these results. By using machine-learning models to computationally predict this ideal substitution point prior to synthesis, the current study expands on this understanding. Machine Learning for Predictive Material Optimization Nonlinear dependencies in materials datasets can be effectively modeled by machine-learning algorithms like Random Forests, Support Vector Regression (SVR), and Gradient Boosting Machines (GBM) (Ward & Wolverton, 2020). While GBM gradually minimizes error through sequential learning, Random Forests combine decision-tree ensembles to improve generalization 405 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) and lessen overfitting. Grain parameters, sintering temperatures, and compositional ratios all work together to influence electrical polarization and dielectric performance in ferrites, and ML can simulate this relationship. For example, Jain et al. (2021) achieved R2 > 0.90 by using machine learning techniques to predict the Curie temperature in perovskite oxides. Similarly, Raccuglia et al. (2016) used AI algorithms to guide experimental synthesis in vanadium oxide systems, significantly reducing trial iterations. These examples show how ML can be transformative when combined with focused experimental design. By using experimental descriptors as inputs for polarization prediction, such as particle size, porosity, lattice parameters, and Nd concentration, the current study adapts such methodologies to R-type ferrites. An optimized composition with improved electrical properties that have been empirically validated is the end result. Conceptual Framework This study's conceptual model combines machine-learning prediction with experimental synthesis in a cyclic feedback loop. ML models use empirical data from synthesized samples as training inputs to forecast ideal synthesis conditions. These predictions guide further experimental validation, forming a self-improving system for material optimization. This hybrid framework promotes theoretical understanding of structure–property relationships while also speeding up discovery. The method is a prime example of a data-driven paradigm shift in materials science, where human expertise and algorithms combine to create innovative solutions that are sustainable, effective, and repeatable. The diagram below presents the integrated workflow of the study. It shows how experimental synthesis, microstructural characterization, and machine-learning prediction are connected in a cyclic feedback system. The process starts with synthesis of Neodymium-substituted R-type hexagonal ferrites, followed by structural and electrical characterization to extract physical parameters (lattice constants, grain size, porosity, macrostrain). These parameters serve as input features for machine-learning models such as Random Forest Regression (RFR) and Gradient Boosting Regression (GBR), which predict the resulting electrical polarization. The predictions guide the optimization stage, identifying the ideal substitution level and processing temperature for maximum polarization with minimal dielectric loss. Optimized results are then experimentally verified and re-fed into the system, creating a closed-loop learning cycle. This conceptual model underpins the entire research framework and demonstrates the synergy between computational prediction and experimental validation in materials informatics. 406 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) Figure1 Conceptual Framework of the Machine-Learning–Driven Design and Optimization Process Explanation The conceptual framework of the current study is shown in Figure 1, which also shows the cyclical relationship between optimization, machine-learning prediction, characterization, and experimental synthesis. The first step in the process is Experimental Synthesis, which uses the sol–gel auto-combustion route to prepare neodymium-substituted R-type hexagonal ferrites (Ba₂Co₂₋ₓNdₓFe₁₆O₂₇) under controlled conditions. Key descriptors like lattice constants, grain size, and macrostrain are then extracted from the synthesized samples using microstructural characterization techniques like X-ray diffraction, SEM, and dielectric testing. The Machine Learning Prediction stage uses these experimentally determined parameters to teach algorithms such as Random Forest and Gradient Boosting to predict electrical polarization from input variables. The optimal neodymium substitution level and processing temperature to maximize polarization are determined by analyzing and using the predictive results for optimization and design. Ultimately, a closed feedback loop that continuously improves material performance is created when the optimized parameters are fed back into the synthesis stage. In line with the Materials Genome Initiative's goal of expedited, data-driven materials discovery, this framework operationalizes the study's hybrid empirical–computational methodology. It illustrates how time, expense, and uncertainty in the optimization of functional materials can be decreased through iterative learning between the experimental and machine-learning domains. Research Methodology Research Design The electrical polarization behavior of neodymium-substituted R-type hexagonal ferrites was analyzed and optimized in this study using a quantitative–empirical design backed by computational modeling. In order to determine predictive relationships between synthesis 407 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) parameters, microstructural features, and functional properties, the study combines experimental data with supervised machine-learning algorithms. Three interconnected stages make up the sequential hybrid approach used in the design: To acquire primary datasets, Ba₂Co₂₋ₓNdₓFe₁₆O₂₇ ferrites (0 ≤ x ≤ 0.25) were synthesized and characterized experimentally. Data preprocessing and feature extraction, which includes dielectric, morphological, and structural descriptor normalization Development and optimization of machine learning models that predict electrical polarization and dielectric loss using the Random Forest Regression (RFR) and Gradient Boosting (GB) algorithms This design creates a strong synergy between artificial intelligence and experimental science by guaranteeing that computational models are validated through laboratory results and trained on empirical data. The study's philosophical position is in line with post-positivism, which emphasizes the natural sciences' objectivity, quantification, and verifiable patterns (Creswell, 2014). The research attains both empirical rigor and predictive flexibility by fusing probabilistic machine-learning inference with deterministic physical theory. Population and Sampling The population of the study consists of synthesized ferrite samples representing diverse compositions within the Ba₂Co₂₋ₓNdₓFe₁₆O₂₇ family.The sampling frame included 420 distinct samples, each corresponding to unique processing conditions (sintering temperature, composition ratio, grain size, porosity, etc.) recorded from controlled sol–gel synthesis. To ensure representativeness and variability, the sampling covered the following substitution levels:  x = 0.00, 0.05, 0.10, 0.15, 0.20, 0.25 Each composition was sintered at three distinct temperatures (1100 °C, 1150 °C, and 1200 °C), generating 18 base conditions. Further, for each condition, multiple replicates (n = 20–25) were synthesized to capture microstructural variability. This approach provided a robust dataset encompassing both compositional and process-induced diversity, suitable for statistical and computational modeling. Sampling was purposive and stratified — purposive in targeting neodymium-substituted R-type ferrites, and stratified by composition and sintering temperature. This approach aligns with the study’s predictive goals, ensuring balanced data across key variables. Data Collection Procedures Synthesis of Samples All samples were synthesized via a sol–gel auto-combustion method, selected for its homogeneity and fine particle control. Analytical-grade precursors (Ba(NO₃)₂, Co(NO₃)₂·6H₂O, Fe(NO₃)₃·9H₂O, Nd(NO₃)₃·6H₂O, and citric acid) were dissolved in deionized water under constant stirring. Citric acid acted as a chelating agent to ensure uniform metal-ion distribution. 414 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) Comparative Performance and Efficiency Table 6 summarizes the comparative improvement achieved through ML-assisted optimization relative to classical design. Method No. of Experiments Required Avg. RMSE Time to Optimization (weeks) Traditional Sequential Experimentation ≈ 60 0.032 10–12 ML-Assisted Design (Proposed) ≈ 20 0.009 4–5 The machine-learning framework reduced laboratory iteration by ~65 % and overall optimization time by more than half, validating H₅ that data-driven modeling accelerates ferrite design. Summary of Findings Descriptive trends confirmed increasing polarization up to moderate Nd substitution (x ≈ 0.15) followed by slight decline. Correlation and regression analyses established strong positive relationships among Nd content, macrostrain, and polarization. Mediation testing revealed macrostrain as a significant partial mediator. Machine-learning algorithms achieved exceptional predictive accuracy (R² ≈ 0.96). Experimental validation confirmed predictive reliability within ±4 % deviation. Optimization efficiency improved substantially, demonstrating practical advantages of AI integration in materials research. Discussion The present study investigated the machine-learning–driven design and optimization of neodymium (Nd)-substituted R-type hexagonal ferrites for enhanced electrical polarization properties. The results offer strong empirical support for the idea that combining artificial intelligence with conventional materials synthesis can significantly speed up the search for and development of multiferroic materials. Integration of Machine Learning with Materials Design In contrast to classical regression, machine learning was able to capture nonlinear dependencies between electrical properties and synthesis parameters. Near-perfect polarization behavior prediction was shown by the Gradient Boosting Regression model (R2 = 0.963), demonstrating the feasibility of data-driven methods in the design of oxide materials. These outcomes align with Butler et al. (2018) and Ramprasad et al. (2017), who emphasized that algorithmic modeling bridges the gap between quantum-level theory and laboratory experimentation. Furthermore, neodymium substitution and microstrain have the biggest effects on polarization, according to feature importance analysis. This supports the physical mechanism by which rareearth doping causes local lattice distortion that improves dipole moment alignment, as suggested by Pullar (2012) and Sun et al. (2022). The interpretability of the model encourages other ferrite families to use it for predictive optimization. 415 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) Effect of Neodymium Substitution Moderate Nd substitution (x ≈ 0.10–0.15) maximizes polarization, as confirmed by experimental and statistical analyses. Lattice distortion and the formation of oxygen vacancies work together to enhance dipole alignment at this level without causing the hexagonal phase to become unstable This result supports that of Iqbal et al. (2023), who found that enhanced Fe–O–Fe bond asymmetry led to comparable improvements in neodymium-doped ferrites. Polarization decreases after x = 0.15, indicating that secondary phases and defect clustering are introduced by excessive substitution. Macrostrain was identified as a significant mediating variable between composition and polarization by the mediation analysis. This confirms the findings of Sharma et al. (2021), who observed that ferroelectric switching efficiency and domain wall motion are directly impacted by microstructural stress. The current findings advance our knowledge of defect-driven polarization in ferrite systems by quantitatively confirming this phenomenon. Comparative Efficiency and Innovation Conventional materials optimization frequently entails a lot of time and money spent on repeated trial-and-error experiments . The ML-assisted framework proposed herein reduced experimental iteration by nearly 65% and optimization time by half, validating the cost-effectiveness of computationally guided experimentation. This efficiency advantage is particularly meaningful for developing-country research contexts where laboratory resources and high-end instrumentation remain limited. Furthermore, the hybrid model offers both predictive accuracy and scientific interpretability, addressing a key concern in AI-driven research—that black-box predictions must be physically meaningful. By correlating feature importance with structural parameters, the study ensured that computational results remain grounded in established materials theory. Comparison with Prior Research While machine learning has been widely employed in semiconductor and perovskite optimization (Ward et al., 2016; Jha et al., 2019), few studies have applied it to hexagonal ferrites due to their structural complexity. The present study thus extends the frontier by demonstrating that even modest, well-curated datasets can yield highly accurate models. It also introduces a localized dataset specifically for neodymium-substituted Ba₂Co₂Fe₁₆O₂₇ ferrites, adding new empirical contributions to the materials science literature from Pakistan’s research landscape. The outcomes further align with the emerging Materials Genome Initiative principles, underscoring how small-scale research groups can integrate informatics tools with conventional synthesis to achieve globally competitive innovation. Theoretical Implications From a theoretical perspective, this research bridges crystal-chemical theory and statistical learning theory. The confirmation of neodymium’s dual role, as a structural modulator and electronic polarizer, provides deeper insight into the structure–property correlation of 416 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) multiferroics. The successful implementation of ensemble models validates their theoretical robustness in modeling complex nonlinearities inherent to solid-state systems. The mediation of polarization by macrostrain introduces a refined conceptual understanding of how atomic-scale perturbations manifest in macroscopic electrical properties—a relationship central to next-generation multiferroic design. Practical Implications Practically, the findings have broad applications in wireless communication, energy storage, and magnetoelectric sensor technologies, where enhanced polarization directly improves performance. The ML framework can serve as a prototype for industry-oriented laboratories seeking to automate materials optimization pipelines. For developing nations, the integration of open-source ML tools offers a low-cost, high-impact pathway toward innovation. The approach democratizes materials design, enabling resourceconstrained research institutions to contribute meaningfully to global technological progress. Conclusion The present study achieved its objectives by empirically and computationally validating the influence of neodymium substitution on the electrical polarization of R-type hexagonal ferrites and by demonstrating the predictive power of machine-learning algorithms in materials optimization. Key conclusions include: Substitutional Optimization: Moderate Nd doping (x = 0.10–0.15) enhances electrical polarization through lattice distortion and defect-dipole interaction. Mechanistic Mediation: Macrostrain acts as a partial mediator linking compositional changes to polarization, reflecting microstructural coupling effects. Model Reliability: Gradient Boosting Regression achieved superior predictive accuracy (R² = 0.963), confirming the validity of data-driven modeling in ferrite systems. Experimental Verification: Predicted polarization values correlated strongly (r = 0.982) with experimentally measured data, confirming practical applicability. Efficiency Gains: Machine-learning integration reduced experimentation time by over 60%, demonstrating a scalable, cost-efficient pathway for materials research. Overall, the research substantiates that machine learning and experimental synthesis are mutually reinforcing—together forming an intelligent loop for rapid, reproducible, and sustainable materials discovery. Recommendations Adoption of Hybrid Frameworks: Future materials studies should routinely combine experimental and machine-learning methodologies to enhance prediction accuracy and reduce resource consumption. 417 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) Data Infrastructure Development: Institutions in developing countries should establish shared repositories of experimental data to support national-level materials informatics initiatives. Model Generalization: Researchers are encouraged to extend the framework to other ferrite compositions (M = Ni, Zn, Mg) to explore the universality of the ML–polarization relationship. Process Automation: Integration of ML prediction with robotic synthesis platforms can further accelerate optimization by implementing closed-loop design cycles. Policy and Training Support: Government and higher education bodies should promote interdisciplinary curricula integrating materials science, data analytics, and artificial intelligence, cultivating the next generation of computational materials scientists. Cross-Disciplinary Collaboration: Future research should foster stronger collaborations between material scientists, data engineers, and computational physicists to create integrated research teams capable of addressing complex multi-scale material phenomena. Such collaboration can enhance dataset quality, model interpretability, and practical implementation of machine-learning frameworks in industrial contexts. Standardization of Experimental Reporting: It is recommended that researchers adopt standardized data-reporting formats (e.g., FAIR and MGI-compliant metadata templates) to ensure consistency, comparability, and machinereadability of published datasets. This would significantly enhance data sharing, reproducibility, and meta-analytic modeling across laboratories. Open-Access Material Databases: Establishing national and regional open-access repositories for ferrite and oxide materials can democratize research by allowing institutions with limited resources to train and validate machine-learning models. Shared databases will accelerate discovery and prevent redundant experimentation. Incorporation of Advanced Algorithms: Future studies should explore deep learning architectures such as Graph Neural Networks (GNNs) and Convolutional Neural Networks (CNNs) for property prediction. These models can capture atomic-level interactions and microstructural hierarchies more accurately than conventional ensemble methods. 418 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) Sustainable Materials Design Perspective: Researchers should integrate environmental and life-cycle assessment (LCA) parameters into machine-learning frameworks to promote eco-efficient materials development. Optimizing ferrite compositions for performance and sustainability simultaneously will align with global sustainable development goals (SDGs). References Agrawal, A., & Choudhary, A. (2016). Perspective: Materials informatics and big data: Realization of the ―fourth paradigm‖ of science in materials science. APL Materials, 4(5), 053208. https://doi.org/10.1063/1.4946894 Breitman, L. (2001). Random forests. Machine Learning, 45(1), 5–32. https://doi.org/10.1023/A:1010933404324 Butler, K. T., Davies, D. W., Cartwright, H., Isayev, O., & Walsh, A. (2018). Machine learning for molecular and materials science. Nature, 559(7715), 547–555. https://doi.org/10.1038/s41586-018-0337-2 Cheng, J., Shen, Y., Wang, X., & Zhang, J. (2018). Structural and multiferroic characteristics of Biscoff₁₆O₂₇ ferrite ceramics. Journal of Magnetism and Magnetic Materials, 446, 84– 92. https://doi.org/10.1016/j.jmmm.2017.09.073 Creswell, J. W. (2014). Research design: Qualitative, quantitative, and mixed methods approaches (4th ed.). Sage Publications. Friedman, J. H. (2002). Stochastic gradient boosting. Computational Statistics & Data Analysis, 38(4), 367–378. https://doi.org/10.1016/S0167-9473(01)00065-2 Iqbal, M., Rehman, A., & Farooq, T. (2023). Electrical and magnetic properties of neodymiumdoped R-type hexagonal ferrites. Journal of Advanced Materials Research, 13(2), 115– 127. Jha, D., Ward, L., Paul, A., Liao, W. K., Wolverton, C., & Choudhary, A. (2019). Element: Deep learning model for predicting materials properties without human intervention. Scientific Reports, 8(1), 17593. Kalidindi, S. R., & De Graef, M. (2015). Materials data science: Current status and future outlook. Annual Review of Materials Research, 45(1), 171–193. Pullar, R. C. (2012). Hexagonal ferrites: A review of the synthesis, properties and applications of hexaferrite ceramics. Progress in Materials Science, 57(7), 1191–1334. Ramprasad, R., Batra, R., Pilania, G., Mannodi-Kanakkithodi, A., & Kim, C. (2017). Machine learning in materials informatics: Recent applications and prospects. npj Computational Materials, 3(1), 54. https://doi.org/10.1038/s41524-017-0056-5 Sharma, P., Gupta, V., & Kumar, R. (2021). Structural, dielectric and magnetic properties of Nd³⁺ substituted Ba₂Co₂Fe₁₆O₂₇ ferrites. Materials Chemistry and Physics, 272, 124948. Shah, F., Khan, M., & Siddique, H. (2020). Synthesis and magnetic characterization of Co-based hexagonal ferrites. Pakistan Journal of Scientific Research, 72(4), 221–228. Sun, Q., Liu, H., & Zhang, D. (2022). Influence of rare-earth substitution on structural and magnetic properties of W-type and R-type ferrites. Ceramics International, 48(5), 6271– 6282. 419 Journal for Current Sign Online ISSN (3006-1504) Print ISSN (3006-1490) Ward, L., Agrawal, A., Choudhary, A., & Wolverton, C. (2016). A general-purpose machine learning framework for predicting properties of inorganic materials. npj Computational Materials, 2, 16028. Ward, L., & Wolverton, C. (2020). Atomistic calculations and machine learning for design of multi-principal element alloys. Current Opinion in Solid State and Materials Science, 24(4), 100797.