Economic Categorizing Based on DFT-induced Supervised Learning

Economic Categorizing is a process of assigning labels to unknown economic events based on known information. In this article, we contrive an algorithms which would serve the purpose of economic categorizing via Supervised learning (SL) on discrete Fourier transform space. SL is a very important approach searching for the relation between feature vectors and labels via a given training set and a test set. If the underlying data is time-dependent or if we are aiming at filtering out some noisy information, then Fourier Transform provides a good technique to achieving such goal. In this paper, I devise a supervised-learning algorithm which converts the raw economic time-series data into frequency domain, measures the distances between test elements and training set, compares their efficiencies and chooses the optimal combination of a method, an approach and a frequency period. This combination would serve our labelling function for economic categorizing of some economic events. Our algorithm could be implemented in machine, and thus such economic categorizing could be enhanced through machine learning. This categorizing algorithm could enrich or supplement the typical classifications. It also provides a dynamical analytical perspectives in classifying.