Auto-dependence Structure Estimating And Forecasting Of Duration Based On Vine Copula

Finance is the core of modern economy. With the deepening financial globalization and the strengthening of synergistic in major economic markets, the era of economic globalization has arrived. Financial risks are inherent properties of finance. Risks go with the existence of financial activities. Global financial integration has led to interdependence between the markets, so that the volatility of local financial market in any region could spread to other financial markets thus becoming a financial tsunami. Over the past decade, the continuous outbreak of the financial crisis, has drawn increasing attention to financial risk management and metrics.As we all know, the stock market is a reflection of the financial situation. In an efficient market, all sorts of information in financial markets will be reflected in the fluctuations in the stock market, at the same time, trading volume is a fundamental measure of the risk status in the stock market Therefore, the study of trading volume, can reveal and explain some of the laws and phenomena in financial markets. Trading Volume Duration refers to the accumulation of time interval in the stock market when generating a certain amount of the transaction. Frequency, intensity, the degree of liquidity as well as the risk level of the stock market can be characterized by the duration of the trading volume. Meanwhile, as trading volume duration changes dynamically over time, with a certain degree of clustering, thus introducing the dependence of the duration to study the concept of trading volume has become a focus and cutting-edge issue, which provided a new way of thinking in assessing and predicting financial risks more accurately. This article is promoted in this background. We proposed a modified model, that is a semi-parametric model based on Canonial vine Copula, to describe and predict the features of auto-dependence in the trading volume duration.At first, this article reviews the status of theory promoted by scholars in related fields of financial risk management, trading volume duration, as well as binary and multivariate auto-dependence model framework, including ACD model and its the expansion models, as well as Copula and vines Copula model.After the pectination of related theories, we pointed out the deficiencies that the traditional methods have in estimating the distribution of multivariate dependence structure. Then we proposed the estimation method of self-dependent structure in the trading volume duration based on of Vine Copula methods.In this paper, assuming all of the existing trading volumes have an impact on the subsequent duration. When analyzing the non-linear dependence structure between duration, we can actually esteem the n adjacent duration as a certain multivariate distribution. We can divide this multivariate distribution into two parts to each study, namely the unconditional marginal distributions of variables and the dependence structure portion where variables link between variables each section. In describing the dependence structure, since every linkage has a dominant node, so the choice of estimation of multi-variable auto-dependence structure in this paper is C Vine Copula. We adopt Copula method to study the trading volume duration high-frequency data. After the estimation of several auto-dependence structure for a series of trading volume duration, we put forward a new conditional density function estimation method, and proposed the forecast of a given trading volume duration.In the empirical part of this paper, the proposed the model for estimating and predicting the duration of the trading volume of China Petrochemical high frequency tick-by-tick data, with EACD (1,1) model as the reference model for comparison. The empirical results show that both EACD model and our model can predict the aggregate effect of the proposed merger duration well, however our model can better predict the duration of the next trading volume, especially when the trading volume duration suddenly increases or decreases. In most cases, this model can be more sensitive in responding financial changes. These findings indicate that the adoption of vine Copula in our semi-parametric model enhanced the flexibility and accuracyt when describing the dependency structure of trading volume duration. Last but not least, we tested the two analyzed models using the density prediction method, the test results received the original hypothesis which says that the cumulative distribution function of the sequence generated in our model abide by uniform distribution, while rejected the original hypothesis which says that the cumulative distribution function of the sequence generated by the EACD model abide by uniform distribution. This also shows that the semi-parametric vine Copula model proposed in this paper also performed better in the density function prediction test.