Structure estimation in Bayesian Networks via scoring and restrict method: an application in the area of agriculture

The need for an assertive predictive analysis combined with the causal interpretation of its result is one of the challenges found in machine learning methodologies. Bayesian Networks methodology proposes to be a tool for causal investigation and is widely used not only for this way but also for other tasks. In this context, this paper aims to find a predictive and interpre-table architecture for the proportion of failures in sugarcane fields using a combined Bayesian Network structure estimation algorithm. Therefore, a hybrid procedure namely scoring and res-trict is proposed as well as its suitability for analysis in artificial and real data is verified. From these analysis, we may conclude that there are an improvement in the adequacy measures and the possibilities of the model interpretation.