Consumer preference in produce, defined by shape and size, heavily influences this market. Understanding the environmental and management factors that impact these features can improve a farmer's economic margins. Since sweetpotatoes are hand-harvested and tend to have varying shapes and sizes, this can result in unpredictable profit margins. Methods for predicting the aesthetic characteristics of sweetpotatoes using environmental and agronomic factors with machine learning have not been developed. Moreover, predicting crop shape and size using agricultural data analysis is challenging due to the need for integrating diverse and complex datasets, including genotypes, weather, field management, and spatial information, into predictive models. This study employed an iterative process involving data preparation, feature engineering, variable selection, and model selection to develop machine learning models that predict sweetpotato aesthetic traits from agronomic inputs. We collected and organized data from various sources with different formats, spatial, and temporal resolutions. After comparing the performance of different machine learning methods using cross validation, Bagging regression had the least predictive error in terms of RMSE and MAE for sweetpotato's length-to-width ratio (RMSE = 0.185, MAE = 0.147) and curvature (RMSE = 0.013, MAE = 0.010) predictions. Bagging regression outperformed a naive baseline by 29%–38% when predicting sweetpotato features Our study also determined that the Covington cultivar and GPS locations were the most important factors that influenced the shape and size of sweetpotatoes. Fertilizer prior to planting, rose as an important feature when predicting sweetpotato curvature. Precipitation had a greater impact on the prediction of length-to-width ratio when compared to predicting curvature. The methodology presented herein could be applied to other crops like cucumbers, eggplants, peppers, and potatoes, where the size and shape are important factors for determining their value.