To optimize the projection function and direction of projection pursuit technique, predigest its realization process, and overcome the shortcomings in long time calculation and in the difficulty of optimizing projection direction and computer programming, this paper presented a modified simplex method (MSM), and based on it, brought forward the eco-value level classification model (EVLCM) of forest ecosystem, which could integrate the multidimensional classification index into one-dimensional projection value, with high projection value denoting high ecosystem services value. Examples of forest ecosystem could be reasonably classified by the new model according to their projection value, suggesting that EVLCM driven directly by samples data of forest ecosystem was simple and feasible, applicable, and maneuverable. The calculating time and value of projection function were 34% and 143% of those with the traditional projection pursuit technique, respectively. This model could be applied extensively to classify and estimate all kinds of non-linear and multidimensional data in ecology, biology, and regional sustainable development.