Study On Prediction Model Of Grain Post-harvest Loss

With the coming of the information age,a lot of data have been generated in a series of processes such as harvest,processing,transportation and consumption of grain.Using these data reasonably,we can get more valuable information from them.Under the support of the national grain public welfare project,this dissertation investigates the postpartum loss and waste of grain,and studies the evaluation technology of postpartum loss by the establishment of user portrait technology and prediction model.User portrait is a tagged user model based on a series of real data and abstracted from users' demographic characteristics,behavior preferences and other information.It can be used for user classification statistics,precision marketing,intelligent recommendation system,service or product customization,business management analysis,etc.Based on user portrait,this dissertation studies the loss data of grain postpartum consumption in multiple links.The specific work is as follows:First of all,we believe that user portrait is the mathematical modeling of users in the real world,and the core is establishing label system.Tags are symbolic representations of some user characteristics,and user preferences can be expressed by tags.Based on the data of grain postpartum loss survey,this dissertation analyzes the characteristics of these data.By analyzing the characteristics of grain data,some text data are processed numerically.At the same time,the method and method of determining grain postpartum loss are expounded.Secondly,we should establish the overall framework of user portrait and design the specific content of user portrait.For non numeric data,TF-IDF algorithm is used to calculate the weight of each attribute.For the numerical data,the Pearson correlation coefficient is used to calculate the tag weight of each attribute of the user.Then label each user and calculate the tag value of the user with the attribute value.On this basis,K-means clustering analysis is used to divide the food user groups.Finally,Based on the user portrait technology,a variety of classification prediction models for grain postpartum data are established by using simple Bias,decision tree,support vector machine and other classification methods.Several groups of comparative experiments were conducted to verify the effectiveness of the division of food user groups.Based on this,a loss prediction model of support vector machine based on user portrait is proposed.We use confusion matrix as the evaluation index,take the accuracy,coverage and hit rate to verify the prediction effect of the model.