Price Forecast Of Pledged Property Based On EMD In Inventory Financing

Inventory financing is an emerging business which is based on stocks as the pledge to reduce or avoid credit risk.In practice, banks need valuate stocks to judge whether it can keep on the financing capability of guaranty in the future. Therefore, it is essential to accurately measure the fluctuation risk of the stocks’price to provide quantitative decisions of the pledge loan ratio.Stocks’price risk is one of the key risk elements in inventory financing, to study stocks’ essential characteristics and fluctuation trends can effectively control the uncertain losses that banks face with in the business. This paper, in view of the complexity and volatility of the stocks’price, based on daily trading price series of Shanghai rebar, tries to decompose the price series into several Intrinsic Mode Functions (IMF) and a residual with Empirical Mode Decomposition (EMD) models, then reconstruct the IMFs into the high and low frequency mode with fine to coarse reconstruction algorithm, and next discuss the fluctuation of each mode and the influencing factors of each component. The high-frequency mode reflects the market short-term fluctuations on the impact of price sequences, the low-frequency mode is a reaction by great events during the period of volatility, and the residual, the most important part of the price series, is the long-term trend formed by Comprehensive games of Market trading stakeholders.On the basis of the above research, this paper selects different parameters combination to forecast three component sequences——high frequency mode, low frequency mode and residual trend with Support Vector Machine (SVM) models based on Lib-SVM toolbox, and uses appropriate integrated prediction method to get the final results. In order to evaluate the model prediction precision, it employs four evaluation indexes of mean absolute percent error(MAPE), root mean squared error(RMSE), squared correlation coefficient(r2) and directional statistics(Dstat) to test the prediction results of two stages, and also compares the integrated results with those of single SVM models.It turns out that, there are many superiorities of describing the nonlinear characteristics of price sequences and seeking the global optimal with the model of SVM based on EMD by means of theoretical modeling and empirical analysis, which provides effective tools for banks to measure the price risk of stocks during the period of pledge.