Study On The Method Of Projection Investment Risk Analysis Based On Support Vector Machine

With the reform opening policy and the initial establishment of market economy in china, especially joined the WTO, investing projects in recent had become more complex and larger, which led to risk happened frequently, therefore, it necessary to provide higher requirements for project management. One of the main problems in current project investment is omitting risk analysis in project investment decision and lacking a practical effect method for complex project investment decision. Hence, it's significance in theory and practice to study risk analysis method for project investment. In addition to facilitate the established and improvement system of risk analysis method in theory, it can also promote the standardization and rationalization for project investment decision in practice and correspondingly improving investment success ratio.The main research works and corresponding contributions of this dissertation are as follows: Firstly, three key and related issues of project risk analysis, i.e. risk identify, risk estimation and risk evaluation, have been detailed analyzed in section two. Some traditional method of risk identify, such as expert analysis questionnaire, fault tree analysis and diagnoses filter monitor etc., has been analyzed, discussed and summarized. Then, coupled with the general characters of investment risk and the relationship between economic appraisement and risk, this paper proposed to analysis risk factors according to feasibility research report, i.e., from technical measurement, organization and management, social environment, finance and economic etc. six aspects to analyze the risk factors of project investment.After pointing out the inadequate abilities of traditional evaluation methods, this dissertation introduces newly development method, statically learning theory for project risk analysis, and proposes a novel data drive model method for projection risk evaluation based on support vector machines (SVM). SVM has been analyzed in theory and the author shows that it has better generalization and stability than BP neutral network with small sample data, which is critical to practical case duo to the facts of troubling, or even impossible to collecting enough training datasets, hence naturally SVM takes more advantages in projection investment risk evaluation. In addition, the proposed method has superior to traditional methods with the characteristics of lower cost, easy to operation and batch process for risk evaluation. To improve the data credibility and precision, it's important for data driving model to...