In order to reasonably reflect farmers pesticide behavior risk factors and relatively accurately predict the probability of farmers pesticide application behavior. The article is based on Bayesian network theory to identify the behavior risk factors of farmers pesticides application，which builds a naive Bayesian network model. Applying data from 986 samples within China，after training and validating the model，the article calculates the posterior probabilities of farmers pesticides application behavior and assesses risk of this behavior. The results show that almost all of the farmers used pesticides. Under some specific circumstances，such as agricultural production and operation types，production and agricultural land scales，and productions’ use etc.，were established respectively，the risk of unplanned spraying intervals probability would be greater. If pre-sale price of products and the pesticides price are established respectively，the fluctuation appears on how farmers mind pesticide residues. Low price of pesticides is the main reason why farmers ignore pesticide residues. In addition，the pre-sale price of agricultural products has a great impact to excessive use of pesticides. Therefore，the government should strengthen the supervision and management of the production of agricultural products and prevent the emergence of issues regarding to agricultural products actively. The authority should guide and encourage farmers using pesticides reasonably. Pesticides application training also should be considered，which helps farmers to conduct pesticides application effectively，so that the quality of agricultural products will be improved.