A HIDDEN MARKOV CHAIN APPROACH TO CROP YIELD FORECASTING

Prediction of harvest yield is an important and challenging problem. Attempts to solve this problem rely usually rely on regression techniques highly dependent on local factors. This paper presents a hidden Markov model approach for forecasting weight production. The model can deal with any culture or provided data. Results show that the model can capture both spatial and temporal harvest variability. Model analysis can help determine causes of variability, differently from regression or more straightforward Markov chain approaches. The resulting structure can benefit from statistical techniques for model tuning and model fitting.