This study aims to analyze research trends and knowledge structures in explainable artificial intelligence (XAI) for predictive modeling through a bibliometric review approach. Data were obtained from Scopus metadata in BIB format exported on June 30, 2026. The initial search strategy yielded 343 documents, which were then filtered based on article document type, English language, and journal source, resulting in 177 articles analyzed for the period 2015-2026. The analysis was conducted using a bibliometric approach through mapping publication productivity, journals and primary authors, influential documents, country contributions, keyword co-occurrence, trend topics, thematic maps, and thematic evolution. The results show that XAI research for predictive modeling experienced a strong acceleration after 2023, indicating a shift in focus from predictive models that are solely accuracy-oriented to models that are transparent, explainable, and accountable. The intellectual structure of this field is interdisciplinary, with contributions from computer science, education, health, energy, environment, geospatial, industry, materials, and engineering. The dominant themes center on machine learning, data mining, forecasting, interpretability, SHAP, LIME, and deep learning, while emerging themes focus on counterfactual explanation, causality-aware forecasting, physics-informed learning, transformers, and graph neural networks. This study identifies five key gaps: method, data, application, theory, and evaluation. The primary contribution of this research is to provide a systematic mapping of the developments, intellectual actors, dominant themes, emerging themes, and future research agendas of XAI to build more accurate, transparent, auditable, and accountable predictive modeling