Navegando por Data de Publicação, começando com "2021-06-20"

Filtrar resultados por ano ou mês

Agora exibindo 1 - 3 de 3

Estudo do teste SNHT (Standart Normal Homogeneity Test) para detecção de pontos de mudança em séries temporais
(Universidade Federal de São Carlos, 2021-06-20) Chianezzi, Giovanna Garutti; Moura, Maria Sílvia de Assis; https://lattes.cnpq.br/9410151859448447; https://lattes.cnpq.br/7655895588516271
The study of time series is extremely important for a better understanding how certain events develop over time. To do this, be able to to indicate the exact moment when a given phenomenon changes its pattern of behavior is a very nice feature. This work goes deeper into the theme of detecting change points in series temporal through the use of the SNHT (Standard Normal Homogeneity Test), which consists of a statistical test proposed specifically for this purpose. Some basic statistical concepts and the test in are only explained in detail. The example of the test is performed by applying it to data about the number of drivers killed or seriously injured in traffic accidents in Great Britain. Brittany between the years 1969 - 1984. An experiment, from series of simulated moving averages, will be carried out for that it is possible to get a good sense of the power of the test and its performance. It is analysis will be done through the analysis of Type II Error rates obtained by SNHT. And to complete the study completely, the SNHT will be applied to data referring to the number of new cases of Covid-19 in the city of São Paulo, in an attempt to understand Covid-19's pandemic behavior during its first year.
Modelo de Séries Temporais Autorregressivo Periódico - PAR
(Universidade Federal de São Carlos, 2021-06-20) Cavalaro, Natália Leite; Moura, Maria Sílvia de Assis; https://lattes.cnpq.br/9410151859448447; https://lattes.cnpq.br/0280180212825579
This work presents the study of a type of time series model, called of periodic autoregressive model, which emerged from the researches of Thomas and Fiering (1962), according to Hipel and McLeod (1994). Its use is mainly in series temporals that present a periodic behavior in the mean, variance and function of autocorrelation. Application examples will also be displayed, in which the time series presents the ideal characteristics of using the PAR model, and how to do the procedure of choice, study, suitability and prediction of this type of model, in addition to being carried out a comparison with the seasonal time series model.
Utilização de aprendizagem de máquina para classificação de e-mails em categorias relevantes
(Universidade Federal de São Carlos, 2021-06-20) Silva, Bruno Ferreira; Levada, Alexandre Luis Magalhães; https://lattes.cnpq.br/3341441596395463; https://lattes.cnpq.br/0410389187604892
One of the main technology tools currently used to exchange information is the email service. However, managing the high volume of information received is one of the major challenges encountered in using this service in public and private institutions. Automated text classification has been considered an essential method to handle a high of textual information that people have to deal with on a daily basis. Problem solving by electronic and automatic means is increasingly common due to the reduction of manual work and costs. With that in mind, companies that receive support requests via email have been trying to reduce service time by using machine learning algorithms to sort texts sent via email. This study aims to identify the ability of machine learning algorithms to correctly determine categories, using a previously labeled database. The results were calculated as means and standard deviations of the most used metrics in machine learning, as well as the execution time of the four algorithms used. The results showed themselves to be satisfactory and well functional.