Norovirus is a leading cause of acute gastroenteritis across all age groups, particularly affecting young children. Given its strong seasonal patterns, accurate prediction of detection rates using climate variables is crucial. In this study, wavelet coherence analysis was employed to explore the time-varying relationships between climatic factors and norovirus detection rates, with particular attention to phase differences in the annual cycle. To account for long-term nonlinear trends, generalized additive models (GAMs) were applied to estimate and remove underlying temporal patterns, thereby enhancing seasonal signals in the data. The adjusted dataset was used to train prediction models on data from January 2007 to June 2019, while performance was evaluated on data from June 2019 to December 2020. Four machine learning models were implemented to predict weekly detection rates, incorporating wavelet-based features that capture dynamic seasonal behavior. The proposed framework demonstrated improved predictive performance, achieving approximately 10--15% higher accuracy compared to models relying solely on raw climate variables.