Using Copula And Stochastic Gradient Boosting To Analyze The Risk Of Mechanical Industries Profit

Mechanical industry is one of backbones for developing in China's 11th five-year plan. Flouring mechanism is necessary to enhance big country to great power. Strengthening heavy duty machinery base construction stares Tianjin in the face with China's developing stratagem to take Tianjin Binhai New Area as third engine. With great investment, poor return and more adventure in mechanism, quantificational analyse of risk in kee-jerk investment project is of great importance. It will be useful for investor's decision to know the probability of exceeding certain value in profit in the future and the adventure of his project.There are lots of adventures in influencing company's gain. Market demand and cost, preferred here, are the two key factors to determine profit. We are choosing the indicator, purchase of equipment and instruments of investment in fixed assets, to represent the demand of China mechanical industries and the most unsteady raw material in cost, iron ore, to reflect the variation of the cost in mechanism. Goodness of fit for univariate, such as test for normality, is applied to identify their cumulative distribution of change rate respectively. The most excellent copula in three Archimedian copula, Frank, Gumbel and clayton, is picked out by copula's goodness of fit using probability integration transform. Parameters, like marginal one and dependence one, are estimated by maximum likelihood estimation. Some kinds of dependence measure, such as Spearman'sρ,Kendall'sτ,Gini'sγ,Blomqvist'sβ,Schweizer and Wolff'sσand their extensions, are numerical calculated. The results show that those coefficients are almost the same and there is some negative dependence between two change ratio of market demand and cost.Firstly pursuing the functional relationship between the change ratio of the sum of gain and tax in China mechanical industries and market demand and cost, Stochastic Gradient Boosting is certified to be the most powerful in support vector, multivariate adaptive regression splines, linear regression and itself from the test results of left samples. Relative influence and partial dependence for market demand and cost along with the sum of gain and tax are obtained.Firstly generating pseudo random numbers from the fitted copula and substituting them in the fitted function, the cumulative probability distributions of the sum of gain and tax in China mechanical industries and Tianjin Tianzhong Jiantian Heavy Industries Company Limited in 2010 are simulated by means of Monte Carlo to compute the risk measurement, such as variance, VaR (value at risk), CVaR (conditional value at risk). The sum of gain and income tax of Tianjin Tianzhong Jiantian Heavy Industries Company Limited in 2010 is conservatively predicted to be 1.294 hundred million (at 20% level VaR), i.e. the probability that it will exceed 1.294 hundred million equals 80%, consequently the probability that it will be under 1.294 is 20%. In order to validate the effectiveness of our method for risk analyse, purchase of equipment and instruments of investment in fixed assets, having great linear relationship with gross mechanical industrial output value, is substituted by it, and also is iron ore price, having great linear relationship with steel price. The rationality for our choice of indexes is obvious for linear relationship. Lastly, development strategy of heave-duty machinery industry in Tianjin is suggested.