Soybean diseases are difficult to diagnose due to the similarity of symptoms and to deviations between indicators of a given disease, as a result of differences in local conditions or over time change. The present paper proposes a novel computational technique, evolutionary support vector machines, that can differentiate between four types of diseases found in Soybean. The diagnosis is based on comparing symptoms displayed by considered plants and those of previouslyly predicted cases. This proposed, highly accurate method provides a way of validating expert decision-making and broadening knowledge, and will assist in the management of these diseases.