The strategy to simulate the cross-pollination rate for the co-existence of genetically modified (GM) and non-GM crops by using FPSOSVR

Abstract Recently, genetically modified (GM) technology has become one of the important biotechnologies in agriculture. However, people still misgive the effect of GM crop products for the health and the environment. Hence, the strategies for the co-existence of GM and non-GM crops are a recognized necessity. Using the fitting function to predict the rate of cross-pollination for determining a befitting distance is one of the strategies for the co-existence of GM and non-GM crops. Although many methods have been proposed to predict the rate of cross-pollination, most of them only use the distance from the pollen source to the pollen recipient. However, the rates of cross-pollination are affected by many variables (factors). Furthermore, the cross-pollination rate prediction might be underestimated because most of cross-pollination rates are close to zero. Accordingly, to solve these problems, this paper proposes a hybrid method (FPSOSVR), which consists of fuzzy logic, particle swarm optimization, and support vector regression, for predicting the cross-pollination rate of pollen recipient from the pollen source to simulate the real situation of GM and non-GM maize. Compared with three existing methods, including Exponential, Log/log, and Log/square, the FPSOSVR, Exponential, Log/log, and Log/square methods yielded the root mean square error values of 5.0100%, 8.1950%, 8.2559%, and 8.1919% for training and 3.0727%, 7.2883%, 7.3736%, and 7.3026% for testing, respectively. Obviously, the fitting ability and prediction accuracy of FPSOSVR are better than those of the existing methods about 30–50%. Furthermore, the correlation coefficient between the actual and the predicted cross-pollination rates of FPSOSVR is also close to 0.9, that is, FPSOSVR can be successfully applied in the co-existence topic of genetically modified (GM) and non-GM crops because it brings a better performance in terms of several statistical indicators.