Support vector regression and its application in trading
2 likes2,622 views
This document discusses support vector regression and its application in trading. It explains that support vector regression finds a hyperplane that minimizes loss, with loss defined as being zero within a small deviation from the hyperplane. A radial basis function kernel is used to map points to a higher dimensional space where data is linearly separable. Support vector regression with an RBF kernel is well-suited for trading applications because stock market data tends to be noisy, and it achieves small loss for both close and far away data points. Support vector regression can be used for outlier detection, regime prediction, classifying news, and identifying bad trading days.
Support vector regression and its application in trading
- 1. info@tworoads.co.in Support Vector Regression and Its Application in Trading
- 2. Support Vector Machine ● Red and Blue points belongs to two different class. ● SVM finds a hyperplane that separates these two classes. ● It maximizes the minimum distance of given data points from hyperplane.
- 3. Support Vectors Hyperplane obtained depends only on few data points (not on all points) known as support vectors
- 4. Support vector Regression It finds a hyperplane such that Loss is minimized. Loss is taken to be zero within small deviation (ϵ) from hyperplane. Here the hyperplane depends on points lying outside ϵ margin
- 6. Kernel Trick Map points to a higher dimensional space where data is linearly separable. Kernel Trick is a way to map points without having to compute the mapping explicitly
- 7. RBF Kernel K(x, x’) = exp (γ ||x - x’||2 ) where || . || is Euclidean distance and γ is negative We can see that if the distance between 2 points is large then this function gives a very small value. Thus this function achieves small loss for far away points.
- 8. High Dimensional Space RBF basically maps the points to a Hilbert space (infinite dimensional space) and thus SVR with RBF kernel finds a hyperplane in that space.
- 9. Why SVR in Trading Stock Market Data is very noisy. There are lots of outlier point in the data. Their loss will be large, and thus these points will change the hyperplane very much. We would like to have small error for points which are close to hyperplane, Also we would like to have small error for points which are very far from the hyperplane. To achieve this, we use a RBF (Radial Basis Function) Kernel function.
- 10. SVR in Trading ● Outlier Detection ● Regime Prediction ● Classify News ● Identify Bad Trading Days
- 11. References ● Support Vector Regression, by Max Welling ● Support Vector Machine by C. Cortes and V. Vapnik ● Support Vector Regression by Debasish Basak, Srimanta Pal and Dipak Chandra Patranabis
- 12. Thanks for coming! Aashay Harlalka aashay.harlalka@tworoads.co.in www.tworoads.co.in info@tworoads.co.in