Aprendizagem automatizada e estatística aplicada: estudo em análise espectral e mercado financeiro

Abstract

The automation of spectral data analysis is a necessity of the project Physics of petroleum in porous media. Through machine learning algorithms (subfield of computer science) it is possible to classify data where the machine is able to learn the correct parameters of classification models of minerals and oils. The method goes beyond, being applicable in different areas. The study of graphs has intrinsic connection in the definition of the architecture of such algorithms and from graphs of minimum spanning trees (MST) we visualize and group the data. We also apply MST in shares of the American stock exchange. Within the financial market we develop statistical tools that describe the movement of stocks and the pricing of American and European options. Statistical and machine learning methods are used in prediction and inference tasks. For inference models, we want to describe the pattern of a data set through a probabilistic model which is the major focus of statistics. Prediction is the ability to correctly sort unfamiliar samples. Predictive models of deep learning are created in such a way that the algorithm finds high complexity patterns that are hard for a human to identify. With all that in mind, the main objective of this work is to build a set of automatic analysis tools based on graphs, machine learning and statistics that can have great applicability in many areas.