A Hybrid Coarse-to-Fine Algorithm Based on Machine Learning for Predicting Economic Crop Yield

The cultivation of economic crops is of great significance in increasing famers' income and promoting rural economic development. The present study focuses on tobacco yield prediction by proposing a hybrid framework based on machine learning algorithms. The framework combines three sources of data, i.e., expert evaluation data, meteorological data, and planting condition data. As a special economic crop, tobacco planting brings benefits to the local economy and alleviate poverty in underdeveloped areas, but must also be strictly restricted. More accurate yield prediction leverages improve yield regulation. The present study proposed a hybrid prediction framework by combing machine learning algorithms and expert knowledge to achieve coarse-to-fine prediction along with the three stages during the growth period. The empirical studies demonstrate that the prediction framework captures the interactive relationship between the policy-makers and farmers. The present framework provides a digital transformation solution for the economic crop production under highly uncertain meteorological conditions.