Prediction and application on the first flowering date of June-bearing strawberry cultivars using Ensemble-based Machine learning with primary temperature values

• The model improved accuracy by integrating multiple single models. • The prediction reduced errors by cultivar-specific minimum growing degree days. • Machine learning with external meteorological data have more accurate flowering. Predicting the timing of the flowering date is crucial for harvest scheduling, fertigation, and greenhouse environmental control decisions regarding cultural practices. To date, several studies have used temperature-based models to predict flowering dates. However, most developed models were imprecise and inaccurate in an environment independent of the experimental site or date. This suggested the necessity for a simple and accurate model that can be applied across diverse environmental conditions. The primary objective of the present work was to predict the initial occurrence of flowering in strawberries by using temperature-based models within a plastic greenhouse environment to overcome the limitations of existing models. Additionally, the study was to develop a cultivation map by predicting the first flowering date (FFD) within the main region of strawberry cultivation. Unlike previous models, our approach incorporates ensemble learning and a transductive learning strategy that partially fits the structure of each test sample (Test-time partial fit), along with standardized external data, enabling predictions in different environments. The results indicated that the model could accurately predict the FFD using temperature data. This model has the potential to offer fundamental data for the prediction of initial flowering timing, as well as being utilized for the purpose of managing cultivation to control the timing of blooming.