The study of citation lifecycles in academic publications is crucial in scholarly research. Many studies use descriptive statistics or regression analyses to forecast citation outcomes, but they often don't fully combine textual data (like titles, abstracts, and keywords) with numerical data (such as impact factors and h-indexes). This research introduces an innovative multimodal model designed to predict early citation trajectories for scholarly articles, addressing this gap. We developed eight models to predict citations from the first to the eighth year based on 2017 data. Our lifecycle analysis shows that the model maintains high performance over multiple years, highlighting its robustness and adaptability. The results underscore the benefits of combining diverse data types for long-term predictive tasks, making our model a valuable tool for researchers and practitioners in Library and Information Science. This model significantly improves our ability to assess the early citation potential of academic papers, making it a valuable resource for researchers and policymakers in academic publishing. Additionally, to thoroughly explore bibliographic data, the study used LDA to investigate the topic distribution of library and information science publications in 2017.
