Network fault diagnosis of SVM based on information geometry

There are many chaos phenomena in traditional network fault diagnosis and monitoring technology, therefore the fault diagnosis results cannot be accurately predicted by the traditional methods. To overcome the disadvantage, first, the algorithm of Chaos SVM is given based on chaos theory and SVM. However, kernel function is difficult to choose in the chaos-data. So information geometry is introduced to magnify local area in order to construct the kernel function in the chaotic environment. Then comparative experiments are carried out by three methods. Finally, the conclusion is addressed that the new method can predict network fault diagnosis much more accurately.