The investing risk comprehensive evaluation using an improved support vector machine algorithm

The electric power projects face the uncertain external environment, they are complex of projects themselves and the ability of the designers, erectors and operators are limited, which make the investing risk evaluation of electric power project becomes a pressing settlement problem. To evaluate the investing risk scientifically and accurately, this paper proposes the multi-level classification evaluating model based on improved support vector machine (SVM), which uses the SVM classification combination in series and introduces the type weight factor and sample weight factor. The model not only solves the shortcomings of small sample, high dimension, nonlinear and local minima in the traditional model, but solves the wrong classification question caused by the number imbalance of training samples and data interference. The investment risk evaluating results of 14 electric power projects in National Power Company show that the model is simple, feasible, and improve the evaluating accuracy and efficiency.