The Application Research Of Projection Pursuit Regression Model In The Stand Merchantable Volume Rate Forecast

To accurately forecast Stand Merchantable Volume rate has a crucial role in the formulation of forestry development plans, budget and estimate of timber production, implementation of the logging quota system, carrying out Forest Resources Assessment, and other forest management activities. The traditional mathematical modeling to solve the problem starts around the research of stem curve. But the multivariate statistical analysis which is common in agriculture and forestry area is difficult to expand application because of the non-linear, non-normal, and small quantities of sample data. With the computer technology and Statistical learning theory and its applications development, use of learning machines can study the issue of new ideas and solutions.Projection Pursuit is a class of new statistical methods of high-dimensional data processing and analysis, the basic idea is to project high-dimensional data onto low-dimensional sub-space, and find the projection that reflects the structure or characteristics of the original high-dimensional data, in order to achieve high-dimensional data research and analysis purposes. The thinking combined with various specific algorithm have many applications in other fields, including data Dimension reduction and regression forecast.The thesis through the Fujian Agriculture and Forestry University Youth Fund project which executed by the author, relies on R software platform, using the theoretical analysis, program design and numerical simulation method of combining, is to establish a Stand Merchantable Volume rate Projection Pursuit regression forecast model of high degree of operational accuracy and efficiency. And by LOO Cross-validation confirms its generalization performance is stable. The Part 1 of the thesis shows the whole technical route; Part 2 is in the Projection Pursuit ideology high-dimensional data principal component analysis dimensionality reduction theoretical analysis and practical application; Part 3 In-depth studies on the Projection Pursuit Regression coupled model, and establishes the simulation forecast model based on machine learning method; In order to test the model superior and Inferior in the similar methods, Part 4 expands the horizontal comparison on the same issue with Artificial Neural Network, Projection Pursuit Regression, Support vector machine, and Multivariate adaptive regression spline which are all belong to supervised learning mechanism model. Analysis and simulation results show that in the issue of Stand Merchantable Volume rate forecast, Statistical learning method is feasible and reliable, and Projection Pursuit Regression model prediction accuracy and the convergence rate compared with the methods of same class has some advantages, Projection Pursuit applications in forestry has considerable practical value and development potential.