The Application Of Bayesian Networks In Land Use Decision Of Farmers

Based on the problem of the changes of farmers' land use decision under the background of returning farmland to forestry,the Bayesian network is built to analyze the farmers' land use decision.It is found that the labor and government subsidy have a relatively large impact on farmers' land use decision.First of all,the PEMD algorithm is used to discrete the survey data of county households in Xian'nangou,and 14 kinds of Bayesian networks are constructed by the BNT of MATLAB software.Meantime,using Bayesian,LL,AIC and BIC these four scoring functions to evaluate the network.It's show that,the MWST+T+K2 network is selected to research the farmers' land use decision because of the highest AIC score,the higher Bayesian score,LL score,BIC score.As well,its running time is also acceptable.Then,four kinds of Bayesian classifiers such as NBC,TANC,MWST and MWST+T+K2 are tested by using HOLD OUT,CV-5 and CV-10 classification tests respectively.It's found that MWST+T+K2 classifier's classification,level is higher by comparing their classification index.But it only has advantages for the first type of data classification and some local optimal problems.At the same time,the joint reasoning of MWST+T+K2 network structure is carried out and the inference results show that the network has some validity.In the end,proposing the HC&K2 algorithm to solve the problem of the MWST+T+K2 network.It is found that the HC&K2 network can distinguish three types of data by comparing HC&K2 network with TANC and MWST+T+K2 classification.Also,its inference results show that its accuracy is consistent with the MWST+T+K2 network,absolute error is less than 0.015.And the HC&K2 can better identify the combination of data that does not exist.The result is that when the subsidy is 1750,or it's set between 0 and 3500,it is suitable for the farmers with less labor to land seed orchard.If subsidy is more than 3500,they tend to choose waste land resources.