Thais Oshiro

Random Forest is a computationally efficient technique that can operate quickly over large datasets. It has been used in many recent research projects and real-world applications in diverse domains. However, the associated literature provides almost no directions about how many trees should be used to compose a Random Forest. The research reported here analyzes whether there is an optimal number of trees within a Random Forest, i.e., a threshold from which increasing the number of trees would… Mostrar más

Random Forest is a computationally efficient technique that can operate quickly over large datasets. It has been used in many recent research projects and real-world applications in diverse domains. However, the associated literature provides almost no directions about how many trees should be used to compose a Random Forest. The research reported here analyzes whether there is an optimal number of trees within a Random Forest, i.e., a threshold from which increasing the number of trees would bring no significant performance gain, and would only increase the computational cost. Our main conclusions are: as the number of trees grows, it does not always mean the performance of the forest is significantly better than previous forests (fewer trees), and doubling the number of trees is worthless. It is also possible to state there is a threshold beyond which there is no significant gain, unless a huge computational environment is available. In addition, it was found an experimental relationship for the AUC gain when doubling the number of trees in any forest. Furthermore, as the number of trees grows, the full set of attributes tend to be used within a Random Forest, which may not be interesting in the biomedical domain. Additionally, datasets’ density-based metrics proposed here probably capture some aspects of the VC dimension on decision trees and low-density datasets may require large capacity machines whilst the opposite also seems to be true. Mostrar menos

Random Forest is a computationally efficient technique that can operate quickly over large datasets. It has been used in many recent research projects and real-world applications in diverse domains. However, the associated literature provides few information about what happens in the trees within a Random Forest. The research reported here analyzes the frequency that an attribute appears in the root node in a Random Forest in order to find out if it uses all attributes with equal frequency or… Mostrar más

Random Forest is a computationally efficient technique that can operate quickly over large datasets. It has been used in many recent research projects and real-world applications in diverse domains. However, the associated literature provides few information about what happens in the trees within a Random Forest. The research reported here analyzes the frequency that an attribute appears in the root node in a Random Forest in order to find out if it uses all attributes with equal frequency or if there is some of them most used. Additionally, we have also analyzed the estimated out-of-bag error of the trees aiming to check if the most used attributes present a good performance. Furthermore, we have analyzed if the use of pre-pruning could influence the performance of the Random Forest using out-of-bag errors. Our main conclusions are that the frequency of the attributes in the root node has an exponential behavior. In addition, the use of the estimated out-of-bag error can help to find relevant attributes within the forest. Concerning to the use of pre-pruning, it was observed the execution time can be faster, without significant loss of performance. Mostrar menos

A identificação corresponde ao conjunto de procedimentos diversos para individualizar uma pessoa ou objeto. O processo de identificação de indivíduos por comparação manual radiográfica é um processo muito lento e que exige altos níveis de habilidade e precisão por parte do especialista humano. O emprego de sistemas computacionais para auxiliar o processo de identificação pode tornar esse processo mais rápido e prático. Assim, neste trabalho visa-se, com a utilização de medidas obtidas a partir… Mostrar más

A identificação corresponde ao conjunto de procedimentos diversos para individualizar uma pessoa ou objeto. O processo de identificação de indivíduos por comparação manual radiográfica é um processo muito lento e que exige altos níveis de habilidade e precisão por parte do especialista humano. O emprego de sistemas computacionais para auxiliar o processo de identificação pode tornar esse processo mais rápido e prático. Assim, neste trabalho visa-se, com a utilização de medidas obtidas a partir de radiografias dos seios frontais humanos, utilizar métodos e técnicas de Aprendizado de Máquina para verificar se é possível estabelecer padrões entre os sexos, auxiliando assim, o processo de identificação, por meio de medidas propostas em literatura específica da área. Os resultados indicam que as medidas propostas são mais adequadas para especialização do que para generalização. Mostrar menos