This study examines research trends in patent data-based forecasting of promising technologies, using academic publications indexed in the Web of Science from 2004 to 2024. To systematically classify and review methodological developments, the study proposes and applies the PATC analytical framework, which includes Preprocessing & Representation, Analysis Algorithms, Technology Insight, and Context to Action. The analysis covers 268 papers, reviewed in depth by period, methodological approach, and network structure.

The analysis shows a sharp increase in related studies after the 2010s, with a significant rise between 2018 and 2024. Early studies mainly used statistical and network-based analyses, while recent research increasingly uses AI-driven techniques such as text mining, topic modeling, embedding, and deep learning. The co-occurrence network centrality analysis of PATC sub-classifications identifies text processing, statistical analysis, and network analysis as key methodological hubs. In addition, the betweenness-centrality analysis highlights the intermediary roles of new technology detection, vacant technology exploration, and dimensionality reduction techniques. These findings show that patent-based forecasting studies are moving from traditional analytical methods to AI-based approaches, leading to greater methodological diversity. This research offers both academic and practical value by providing an empirical basis for the development of technology forecasting research and for the creation of national policy and corporate R&D strategies.