A New Approach for Fuzzy Fertilization Forecast Based on Support Vector Learning Mechanism

This paper presents a support vector machine (SVM) algorithm to generate fertilization fuzzy rules so as to realize the prediction method. That SVM is equivalent to the fuzzy rule-based modeling (FRM) is important to many practical and complicated situations where one unable to determine the number of rules in advance, such as knowledge acquisition form samples. Agricultural fertilizer commonly used orthogonal experiment whose balanced scattered data makes the regression curve fitting method ineffective. This paper presents training data set for SVM learning, and takes advantage of membership to extract fuzzy IF-THEN rules. Rule activated by the threshold value and credibility controls the prediction process. The approach not only avoids the error caused by regression method, and the fuzzy rules also increase the linguistic interpretability of the generated rules, and improve the capability of knowledge acquisition greatly. The performance of the proposed approach is compared to the linear regression method by orthogonal experiment.