Research And Application Of Prediction Of Porcine Reproductive Production Based On Regression Analysis

Now that the Internet has permeated into all walks of life in society,information management system is constantly improving the efficiency of agricultural production.Data mining technology tends to become more and more mature,and a lot of classical algorithms have been accumulated in practice.However,there is no precedent of data mining implement in the field of pig reproduction and parturition prediction.Therefore,this article studies the model of pig reproduction and parturition prediction by means of data mining,in order to pave the way of improving production efficiency of the pig farm.Sow breeding efficiency is closely linked with and corporate profitability.According to the definition,PSY(Piglets weaned/Sow/Year)refer to the number of weaned piglets the sow can provide per year.PSY level directly affect whether the pig farm's economic efficiency is good or not.In order to predict the PSY accurately,and seek method to improve production efficiency,this article has completed the following work:Firstly,feature selection is conducted based on the actual production experience and historical data.After comprehensive consideration of historical data and actual production,determined features include:pig farm ID,gestational age structure,delivery times,number of born alive,parturition rate of mating,and mortality rate of lactation.Combined with the classical formula of pig reproduction and parturition field,calculate the annual PSY,parturition rate of mating,and lactation mortality rate per pig.Preprocess the data,which includes data structure processing and data text processing.Secondly,the problem is abstracted as a multiple regression analysis problem.Varieties of regression analysis algorithms were used to train the data set samples.The optimal regression analysis was carried out by using the least squares method,BP neural network,M5P,support vector machine,genetic algorithm,random forest and other algorithms.After building regression analysis model to predict PSY,a multi-predictor fusion strategy was proposed based on the basic regression algorithm above.Thirdly,experiments were conducted to investigate the influences of gestational age structure,lactation mortality rate,parturition rate of mating and other index on PS Y.The PS Y predicted values of different gestational age structures,lactation mortality rate and parturition rate of mating were compared,and the results showed that by optimizing the sow gestational age structure or reduce the mortality rate of suckling piglets or improve the parturition rate of mating can significantly improve the PSY.Thereafter,specific measures to reduce the mortality rate of suckling piglets or improve the parturition rate of mating are presented.In regard to the problem of the missing value of the gestational age structure in sample parameters,the regression model and the neural network algorithm are combined to fill the missing value.The computed result shows that the maximum error of the missing value is 3.6%,which is within the required range.Fourth,the results of this paper will be applied to the actual production.The system solves the key problem of the number of weaned piglets which can not be predicted in the production process.It has been well received by users and validated the validity of the model algorithm and software system through practical application.