A Prediction Of Stock Trading Signals Based On Interval Turning Points

The prediction of stock trading signals is one of most popular topic at stock prediction field.Recently,a method combined piecewise linear representation and weighted support vector machine(PLR-WSVM)has shown good performance in the prediction of stock trading signals.Meanwhile drawbacks of PLR-WSVM exist particularly in a real world setting.For example,the performance of PLR-WSVM is unstable because of lack of trading signals,and it is not reasonable to specify same threshold value for all stocks in PLR.In this paper,we conduct a set of improvements to PLR-WSVM.Firstly,we reconstruct input variables,most of absolute technical indicators in input variables are substituted with relative indicators since the relative indicators are generally more helpful in predicting trading signals.Secondly,we propose the concept of interval turning points(ITP).We generate core turning points(CTP)by PLR,then interval turning points are generated according to the range of adjacent core turning points.The interval turning points improves the problem of class imbalance,and also provide more valuable information for classifier.Thirdly,a procedure of selecting a threshold in PLR is provided.The threshold is automatically selected by a given percentage of turning points in a training set.The percentage of turning points is easier to understand by investors compare weigh the threshold.In addition,this paper integrates prior tendency knowledge to identify either buying or selling signals from predicted turning points,which improves accuracy of trading signals identification.At last,we propose an investment strategy with correction function to reduce the possibility of loss.This paper conduct sets of experimental study,and the result confirm the expected performance of improved PLR-WSVM.Besides,the improved PLR-WSVM show steady profits in the long time data sets as well as good generalization on new data sets.