Research On The Diagnostic Methods Of Major Diseases In Pigs Based On Bayesian Network

In the 21 st century,as the world’s largest producer and importer of pork,China has been suffering a sharp decline in pig production capacity under the dual impact of African Swine Fever(ASF)and COVID-19.On the basis of ensuring the "stable production and supply" of the pig industry,the government has successively issued a large number of policies and regulations and vigorously supported the pig breeding industry to build large-scale farms and increase the pig production.With the vigorous development of the Internet,cloud computing and big data,intelligent animal husbandry has become a general trend.In the process of daily pig breeding,farmers only rely on their breeding experience without sufficient knowledge reserve and learning environment,and have no basis to solve the problems encountered in breeding.In addition,the symptoms of pig diseases are complex and diverse,and the traditional disease diagnosis relies too much on the experience of experts,resulting in low efficiency of diagnosis.At present,how to solve this series of problems has become a research hot spot.Bayesian network is an algorithm with simple principle and clear logic,and it is practical and effective in the application process.According to the characteristics of pig disease structure,this paper summarize the law of pig disease by using data resources,select principal component analysis method to explore the disease information,and combine with the Bayesian network algorithm to study the diagnosis model of pig disease,so as to improve the diagnostic efficiency and facilitate the use of farmers.First,collect the raw data,including the pig disease data collected from the pig farm and the information disclosed on the Internet.In order to standardize these data,classify them,label the part of speech and tag the unnecessary data information.With the assistance of authoritative experts in this field,the data information is digitized and normalized,and finally the data pre-processing is completed.Secondly,the preprocessed data is input into the Random Forest algorithm diagnosis model,the Ada Boost algorithm diagnosis model and the XGBoost algorithm diagnosis model,and the model is adjusted and optimized.Then,the pre-processed data are extracted through principal component analysis to extract comprehensive indicators,reduce the data dimension,and improve the advantages of Bayesian network,and then a diagnostic model based on the combination of principal component analysis and Bayesian network algorithm is established.Then the four models are compared and analyzed.By comparing the learning curve and evaluation index of the current four pig disease diagnosis models,it is shown that the diagnosis model based on principal component analysis and Bayesian network algorithm can more accurately judge the disease of pigs according to the symptoms of pigs.Finally,the pig disease diagnosis system is designed according to the characteristics of pig diseases.According to the pig disease information input by the user,the type of pig disease and the probability of the disease are diagnosed.The functions realized by this model mainly include the information retrieval function of pig diseases,the dynamic function of the pig industry and the intelligent diagnosis function of pig diseases.The simple interface allows farmers to make independent diagnosis,which basically meets the daily needs of users and achieves the expected design effect.