Investment risk evaluation of high-tech projects based on random forests model

Abstract

The investment of high-tech projects is characterized with high risk and huge profit, so a scientific and accurate risk evaluation is of importance to make decision. Based on the analysis of the indicators influencing the investment risk, a typical sample set was chosen according to the quantitative score results from the experts. After training, a random forests regression model was established to comprehensively evaluate the investment risk. The assessment result from a case shows the model not only can give a correct rank to the unknown samples, but also give the contribution degree and importance of the indicators. Subsequently, aimed at the leading model parameters, many tests were done to illustration their impact on the output results of the model. Finally, the importance of variables was analyzed, and was compared with that of projection pursuit model. The above results show it has stronger adaptability and robustness. © Springer-Verlag Berlin Heidelberg 2013.