This article proposes a comparison of performance between automatic clusters methods, Kmeans and Partitioning Around Medoids (PAM), applied in simulated data with and without outliers. For this analysis, it’s suggested the decision diagram procedure to define the number of groups, assessing the significance of the variables and the definition of the elements that will compose each group. Whereas both cluster analysis techniques are powerful tools in segmenting the observations, the choice of these tools usually are related to the origin of databases, research costs and knowledge of the researcher on the subject. In the results presented in this paper, the techniques K-means and Partitioning Around Medoids had similar performances with the sample data without outliers. However, when analyzed with the outliers, the Medoids Partitioning Around technique proves to be more robust and consistent compared with the K-means technique
This article proposes the use of X-13ARIMA-SEATS to seasonally adjust the business tendency survey time series of the Brazil Business tendency surveys (manufacturing) (BBTS-M |FGV/IBRE). The BBTS-M is an aggregated time series composed of six indicators, and the article discussed the Direct and Indirect method to figure out the best way to perform the seasonal adjustment. As a result, we observed that although the indirect method had a better performance to capture seasonal patterns in each time series that composes the main indicator, the direct method showed up more stable seasonal factors during the time
Sampling procedures in time to evaluate mosquitoes infestation and other disease vectors must take into account issues of practicality, efficiency and economy of resources. This paper describes an application of the methodology for obtaining optimal design time in Aedes aegypti infestation studies. Starting from an underlying Gaussian process assumption one can develops a decision criterion based on maximizing expected utility allowing penalize or reward the sampling points candidates according to its predictive power and its related infestation risk's outbreak. The inference procedure follows the Bayesian approach and is applied to LIRAa infestation's data for the years 2005 to 2009 in Rio de Janeiro. The methodology has proved to be very flexible with great potential for Health Agencies or Government's Agencies that need to optimize their material and financial resources
The increasing number of applications running on the internet has motivated the study of resilience in computer networks. Researchers and experts aims the standardization of metrics for correct measurement of the resilience level in networks, and this article proposes an indicator of resilience pursuing a good infrastructure from the beginning of the project. This project will be made viable using reliability theory, whose objective is to increase the reliability of network components, reducing costs of maintenance and redundancies, resulting in savings for the project. The proposed indicator was tested in backbone topologies, indicating good results, which allows network managers to use this methodology in the proper planning, in addition to the recovery of any fault incidents. The use of more trusted links will provide a structure better prepared to oppose the threats, and to foment the gain of the quality of services provided by the network, and to comply with Service Level Agreements
