A hybrid forecasting algorithm based on SVR and wavelet decomposition

We present a forecasting algorithm based on support vector regression emphasizing thepractical benefits of wavelets for financial time series. We utilize an e ective de-noising algorithmbased on wavelets feasible under the assumption that the data is generated by a systematic pattern plusrandom noise. The learning algorithm focuses solely on the time frequency components, instead ofthe full time series, leading to a more general approach. Our findings propose how machine learningcan be useful for data science applications in combination with signal processing methods. The timefrequencydecomposition enables the learning algorithm to solely focus on periodical components thatare beneficial to the forecasting power as we drop features with low explanatory power. The proposedintegration of feature selection and parameter optimization in a single optimization step enable theproposed algorithm to be scaled for a variety of applications. Applying the algorithm to real lifefinancial data shows wavelet decompositions based on the Daubechie and Coiflet basis functions todeliver the best results for the classification task