基于支持向量机集成的高新技术企业财务预警分析-financial early warning analysis of high-tech enterprises based on support vector machine integration.docxVIP

哈尔滨工业大学经济学硕士学位论文 哈尔滨工业大学经济学硕士学位论文 - - II - Abstract The high-tech enterprise and industry development is the strategic high ground of todays world economy, science and technology competition. High -tech enterprises are established on the basis of scientific and technological innovation. High-tech enterprises invest a lot and yield huge profit. However, market competition is fierce, the uncertainty of economic policy and the acceleration of technological upgrading directly or indirectly lead to financial risks. Building a model to predict financial crisis can guard against financial risks and protect financial health. Financial crisis is the explosion of financial risks. Building a prediction system can help monitor the productive activities, inform potential risks and make the decision. The article studies the influncial factors and transmission mechanism of financial risks. Finally we choose 18 prediction indicators. The third chapter builds SVM Ensemble model. SVM method has the advantage to classify small sample size, high dimensional, non -linear characteristics of the data, which displays a unique superiority and a good prospect, thus helping to predict financial trouble of high-tech enterprises. By reselecting the training set, bagging integration increases classification accuracy and improves generalization ability. Existing weak classifier fusion techniques only consider the credibility of the classification accuracy, but not take full advantage of the sample data. To solve these problems, we propose evidence-based fuzzy density calculation method. To measure the credibility of the SVM output, we consider 3 aspects: the output credibility of sub-classifiers; the credibility of sub-classifiers; the differences between sub-classifiers. The fourth chapter carries out an empirical test to prove the accuracy of the proposed model and the efficient use when it’s put into practice . On the basis of previous research, we choose the data of 2007 ST high -tech enter