Multi-layer Chaotic Data Analysis: A Scalable Way in Processing Chaotic Data Based on Feature Engineering and Gradient Descent

Chaotic dataset is one of the most common form of dataset, may getting in case of prediction events in the real world. And an important factor to consider is the countless parameters related to a chaotic dataset. This paper develops multi-layer chaotic data analysis, a method to set several layers in order to decrease the amount of parameters in regression algorithms used over chaotic dataset, and make comparisons with traditional method which fits "normal" dataset well. The theory shows a new way to combine human experience with machine learning algorithms, and comparison results show a huge boost of accuracy in practical situation.