This study explores the applicability of generalized machine learning models for predicting carbon emissions across multiple crops at a single location in Iran. Thirteen agricultural crops including alfalfa, beans, cabbage, carrots, corn, cucumbers, irrigated barley, irrigated wheat, lentils, onions, sugar beets, tomatoes, and watermelons are investigated. To enhance the dataset, Conditional Tabular Generative Adversarial Network (CTGAN) data augmentation approach is performed, generating 1000 sample points for each crop. Various machine learning models, including K-nearest neighbors (KNNs), random forest (RF), Lasso regression, multiple linear regression (MLR), neural network regression (NNR), and a novel hybrid model; recursive feature elimination with heuristic nearest regression (RFE-HNR), are employed individually for each crop and collectively in a combined data model. The crops are categorized to facilitate carbon factor prediction model for category of crops. Results are compared with baseline Cool Farm Life Cycle Assessment tool. RF models consistently performed better on different combinations of datasets. This work provides valuable insights into the performance of diverse models in managing complex agricultural datasets and underscores the potential of data-driven approaches in optimizing emissions, thereby contributing to sustainable agricultural practices.