An Integrative Approach Combining Computational Physics and Machine Learning in Complex Nonlinear Systems

Abstract

This study aims to analyze the integrative approach that combines computational physics and machine learning in understanding complex nonlinear systems. The issue examined in this study concerns how researchers build, validate, and interpret computational models that rely on both physical laws and data-driven learning. This study employed a qualitative exploratory case study design involving researchers, lecturers, postgraduate students, and computational practitioners with experience in computational physics, dynamical systems, machine learning, or scientific machine learning. Data were collected through semi-structured interviews, limited observation, and documentation of modeling workflows, simulation records, algorithmic notes, and research reports. The data were analyzed using reflexive thematic analysis. The findings identified five main themes: integrative modeling as a response to nonlinear complexity, negotiation between physical laws and data-driven patterns, validation as a layered scientific process, institutional and technical conditions shaping model development, and scientific responsibility in relation to interpretability and transparency. These findings show that the integration of computational physics and machine learning is not merely a technical procedure, but an interdisciplinary knowledge practice involving theory, data, algorithms, uncertainty, and human judgment. The study contributes to the qualitative understanding of scientific machine learning by emphasizing the role of physical coherence, layered validation, and responsible interpretation in complex nonlinear systems. The findings imply that researchers, institutions, and policymakers need to strengthen transparent modeling practices, interdisciplinary training, computational infrastructure, and ethical standards for using machine learning in scientific inquiry.