Comparing Causal Loop Diagram Algorithms with Machine Learning Approaches for Enhancing Sustainable Performance in SMEs

This study advances the understanding of sustainable performance in Small and Medium-sized Enterprises (SMEs) by moving beyond traditional associational models to infer potential causal mechanisms. We propose a comparative framework that integrates a Causal Loop Diagram (CLD) with advanced machine learning techniques, specifically Causal Forest, and uses Artificial Neural Networks (ANN), and Long Short-Term Memory (LSTM) networks to analyze the interplay between frugal innovation, organizational capabilities, and sustainability outcomes. Using survey data from 236 craft-based SMEs in East Java, the research employs Causal Forest to estimate heterogeneous Average Treatment Effects (ATE) and uncover putative bidirectional causal relationships, which warrant future longitudinal validation. The results reveal several statistically significant reinforcing feedback loops, Learning and Integrating Capabilities strengthen each other with ATE value 0.654 and 0.771, Knowledge Sharing and Application operate reciprocally ATE 0.354 and 0.289 respectively, Coordinating Capability and Environmental Sustainability are mutually supportive with ATE of 0.295 and 0.214, and Frugal Innovation and Integrating Capability form a key loop as shown for the ATE value of 0.174 and 0.313. A reciprocal relationship was also found between Social and Economic Sustainability, with ATEs of 0.283 and 0.303. In the predictive modeling comparison 5-Fold Cross validation is used to avoid overfitting,while the ANN significantly outperformed the LSTM, achieving high R2 scores of 0.9581 for Frugal Innovation, indicating its superiority in modeling complex, non-linear relationships in static data. The study concludes that SME sustainability is driven by interconnected systems of innovation and capability, rather than isolated factors. The synergistic application of causal inference and predictive modeling provides a powerful toolkit for deriving data-driven strategies, enabling SME managers to prioritize reinforcing feedback loops for enhanced resource efficiency and long-term resilience.