Study On The Stand Merchant Ratio Prediction Model Base On Artificial Neural Network

The stand volume ratio is influenced by several facters. such as the mean tree diameter,mean tree height,site,age,the reserved density,volume and so on.The change of merchant volume ratio has the non-linearity and the non-definite characteristic.The forecast of the traditional stand merchant volume ratio uses the method of statistical analysis, needs massive types wooden unit (more than 100 samples).The model has many many parameters which lacks the mature determination method.Artificical Neural Network is a method which based on the example.It dose not need to consider the internal structure of math model, the supposition prerequisite, and the artificial definite of factor weight.As a black box,it comprehensivly maps the integrity of the research object.The ANN has the massively parallel operation, auto-adapted, auto-studied, the fault-tolerant ability which make it become a extremely noticeable new method.and is widly used by biology system which build a model to forecast the non-linearity action. Therefore, the utilization of artificial neural network to forecast the stand merchant volume ratio has the unique superiority, also has provided a new mentality for the forecast of the stand volume ratio.The article take Matalab7．0 as a calculateing platform,apply Artificical Neural Network to build (BP) neural network forecasting models,such as the stand volume forecasting model,the non-specification merchant volume ratio forecasting model,stand diameter class merchant volume forecasting model which all have a three forward feeds directions dissemination. According to the study of the cutting area design material and the actual production code odd number,take mean tree diameter,mean tree height,the reserved density,volume as the input neuron, analy the factors which affected the study efficiency and the forecast precision of the BP network, optimize the BP network model mainly from the quantity of concealment level neuron, the training number, the concealment level excitation function, the quantity of study sample,and build the models.Using the improvement BP neural network to forecast models,the result is: the recognition rate of the stand volume forecasting is 95．4%, the recognition rate of the non-specification merchant volume ratio is 86．2%, the mean absolute error of the stand diameter class merchant volume is smaller than 2%. The result indicated that,Under the small sample condition, the BP forecast model which established has the satisfying fitting precision and the forecast precision.