Petroleum future price prediction method of representative sample-based online support vector regression machine

Abstract

The invention discloses a petroleum future price prediction method of a representative sample-based online support vector regression machine. The petroleum future price prediction method of the representative sample-based online support vector regression machine comprises the following steps of (1) collecting petroleum future price historical data to form a training set; (2) computing a representative sample set of the training set; (3) training the representative sample set through a least square vector regression machine to structure a price predicting model; (4) collecting current trading day data in real time to form a sample to be predicted of xtest; (5) predicting future 3-day average price of the xtest through the price predicting model acquired in the step (3); (6) determining whether the xtest is a new representative sample , if so, updating the price predicting model, and if not, not updating the price predicating model; (7) when the number of samples in the representative sample set reaches a set threshold, simplifying the representative sample set; (8) if no new data arrives, performing a waiting process, else, turning to the step (4). The petroleum future price prediction method of the representative sample-based online support vector regression machine achieves future price prediction through incremental learning, thereby solving the problem of no model updating during long-term data accumulation and improving precision of petroleum future price prediction.

Description

technical field [0001] The invention relates to the field of oil futures price forecasting, in particular to an oil futures price forecasting method based on an online support vector regression machine of representative samples. Background technique [0002] As an important way of oil trading, oil futures has important theoretical and practical significance for managers to correctly control the futures market by analyzing the characteristics of oil futures prices and grasping the fluctuation rules of oil futures prices; for investors to designate correct investment strategies, Avoiding risks and improving returns has an important guiding role. [0003] At present, oil futures price prediction methods are roughly divided into empirical prediction methods, fractal theory methods and machine learning methods. Because there are many factors that affect oil futures prices, and the law of futures prices is nonlinear, machine learning methods have attracted more and more attention...

AI Technical Summary

Problems solved by technology

However, as the size of the data set continues to increase, the training time will increase greatly, and eventually the speed of model learning cannot keep up with the speed of data update.

Therefore, such methods are difficult to operate in practical applications.

Images