Navegando por Data de Publicação, começando com "2018-11-09"

Filtrar resultados por ano ou mês

Agora exibindo 1 - 2 de 2

Algoritmos genéticos multiobjetivo para classificação hierárquica de elementos transponíveis
(Universidade Federal de São Carlos, 2018-11-09) Pereira, Gean Trindade; Cerri, Ricardo; https://lattes.cnpq.br/6266519868438512; https://lattes.cnpq.br/3765060702388111
In Machine Learning (ML), commonly a classification problem consists of associating an instance to only one class within an usually small number of classes. However, there are more complex problems involving dozens and even hundreds of classes arranged in a hierarchical structure, which are known in the literature as Hierarchical Classification (HC) problems. In these problems, an instance is assigned not only to one class but also to its superclasses, and two approaches called Global and Local are often used in HC. In the Local Approach, multiple classifiers are trained using local information from classes, while in the Global Approach a single classifier is induced to deal with the entire class hierarchy, which makes it more interpretable. One of the fields of greatest application of HC is Bioinformatics, where tools that explore hierarchical relationships in data and/or made use of ML are still scarce. In addition, an issue that is important for both Bioinformatics and HC fields, and which is not always given due attention, is the interpretability of the models. In the context of Bioinformatics, a topic that is gaining attention is the study and classification of Transposable Elements (TEs), which are DNA fragments capable of moving inside the genome of their hosts. According to recent research, TEs are responsible for mutations in several organisms, including the human genome, which guaranteed them the nickname of great responsible for the genetic variability of species. In this work, three global methods based on Genetic Algorithms that evolve classification rules applied to HC of TEs were proposed and investigated, and in two of them, Multi-Objective Approaches were implemented in order to better deal with the predictive performance and interpretability objectives. The first one is a traditional optimization method called Hierarchical Classification with a Genetic Algorithm (HC-GA), which was used as the basis for the development of the others. The second method is called Hierarchical Classification with a Weighted Genetic Algorithm (HC-WGA), and implements the Weighted Sum Approach. The third method is called Hierarchical Classification with a Lexicographic Genetic Algorithm (HC-LGA), which follows the Lexicographic Approach. Experiments with the TEs class taxonomy developed by Wicker et al. (2007), have shown that the proposed methods have achieved better or competitive results with the state-of-the-art HC methods from the literature, with the advantage of generating interpretable rule models. When compared to the popular global method Clus-HMC, the proposed methods presented better predictive performance in addition to producing less rules with a fewer tests. In the comparisons with the homology tools BLASTn and RepeatMasker, the hierarchical methods achieved superior results in both datasets and were able to classify all the instances, different from what occurred in those tools. Moreover, it was verified that the two multi-objective methods not only obtained the best results for the two datasets used but they also surpassed the simple optimization method with statistical significance.
A aprendizagem de catálise enzimática por meio de casos: proposta para o ensino de química
(Universidade Federal de São Carlos, 2018-11-09) Goulart, Thiago Pereira; Marques, Clélia Mara de Paula; https://lattes.cnpq.br/0451284088402660
The fundamental aim of provoking innovative forms of knowledge construction in Chemistry during basic education requires the instrumentalization of strategies that can show re-significances of pedagogical practices and understandings about learning in Chemistry, as well as in the other Natural Sciences. Thus, the present study started from the foundations of the strategies of resolution of Investigative Cases and Reflexive Portfolios in order to propose a situation of learning from the enzymatic chemical reactions. The paper proposed the resolution of an Investigative Case written for the specific purpose of which Secondary School students of a private college, approaching a situation related to the daily life of a student of the same series. The writing of the case sought to value the construction of the empathy of the students who would solve it, as a way to dynamize the identification and construction of meanings by them regarding the problem situation presented in the case. The resolution steps were recorded in Reflective Portfolios, and from the totality of these reports produced the body of data used to analyze the reports was extracted. The work of orientations, discussion and resolution of cases involved individual and work steps and group discussions. The main objective of the present study is the characterization of ways in which the development of strategies that value the autonomy, the reflexive character and the identification of the learning objects within the daily life of the students can be possible. From the analysis of the results, four analytical categories emerged that characterize this strategy: the identification of the enzymatic reactions in the daily life of the character of the case; a progressively complex discussion about the structure of the substances and the chemical reactions in which they participate; the identification and discussion of the energy distribution characteristic in determining the occurrence of specific chemical reactions; and, finally, an analysis of the way in which critical judgments are made by the subjects of learning about the sources and the quality of the information they use to support their discourses. Based on the analyzes carried out, it is intended the formulation of a guide with guidelines for teachers and students on how to use investigative cases and reflexive Portfolios in their learning strategies, which concludes the study. One of the possibilities is to join the characteristics, data and discussions present in the Study, elements for an epistemological foundation regarding how it may be possible to raise knowledge in Chemistry from the two instruments presented in the Study - the Cases and the Portfolios - teaching practices that allow expanded views on knowledge, encompassing several areas and different forms of language. The study was carried out in a step-by-step manner, with each one of them orienting the teacher-researcher to the students' own demands. In this way, it was possible to describe learning paths and mechanisms from the forms of reports and reflection presented in the Reflective Portfolios.