The Research On The Prediction Of Forest Resources With Missing Data

Forest planning programming is based on the real and reliable data on forest resources in sustainable forest management. While, the missing data of the forest resources is a widespread problem during the field surveys of all types of forests. In this paper, I use the state-space models to research the complete and the missing data of observation with input variables. First, I presented a brief introduction on the EM method and showed the simulation of EM algorithm by applying linear regression model which also analyzes the reasonableness of EM algorithm. Second, introduce the state-space model and derivate the formulae with forward filter and backward filter in detail, and gives some prediction equation. Third, study the state space model under complete observational data, estimate the true parameters using EM algorithm and, further more we use smoothing to analyze the data. The results between the sequences of smoothing value and the sequences of the observation that generated by the model almost are the same using random simulation. This method is effectiveness. Finally, study the state space model under missing data that is based on complete data sets, and the element of observation are randomly censored with a 15% probability, use the same EM algorithm to estimate the true parameters. From the analysis of smoothing and prediction, the sequence of the value of smoothing and prediction are very much in line using random simulation, thus this method is effectiveness again.