Methodology for Classifying Diseases in Plants Using Convolutional Neural Networks

Convolutional neural networks (CNNs) are one of the deep learning techniques that, due to the computational advance of the last few years, have leveraged the area of computer vision, allowing substantial gains in the most varied classification problems, especially those involving digital images. In this context, this paper aims to propose a methodology for the classification of multiple pathologies related to different plant species. Initially, this methodology involved the image processing and the generation of ten new databases, varying between 50 and 66 classes with greater representation. After training the models (VGG16, RestNet101v1, ResNet101v2, ResNetXt50, and DenseNet169), a comparative study was conducted based on widely used classification metrics, such as test accuracy, f1-score, and area under the curve. To attest the significance of the results, Friedman’s nonparametric statistical test and two post-hoc procedures were performed, which demonstrated that ResNetXt50 and DenseNet169 obtained superior performances when compared with VGG16 and ResNets.

FAKTOR-FAKTOR YANG BERHUBUNGAN DENGAN KEMISKINAN DI DESA JATI MULYO KECAMATAN JATI AGUNG KABUPATEN LAMPUNG SELATAN

This research aims to determine poverty level, and factor related to the poverty level in Jati Mulyo Village, Jati Agung District, South Lampung Regency. This research was conducted in March-April 2018. Respondents were 70 households of poor families. Data were collected with a survey method using quantitative descriptive analysis and nonparametric statistical test (rank Spearman correlation). Results of this research showed that annual expenditure per capita was equivalent to the value of 233 kilograms of rice or the households were included in very poor category. Factors that were significantly related to poverty level were level of education and main occupation, while the number of family members and the age of the head of the household were not significantly related to poverty level.Key words: household, poor, poverty level

Big Data Characterization of Learner Behaviour in a Highly Technical MOOC Engineering Course

MOOCs attract a large number of users with unknown diversity in terms of motivation, ability, and goals. To understand more about learners in a MOOC, the authors explored clusters of user clickstream patterns in a highly technical MOOC, Nanophotonic Modelling through the algorithm k-means++. Five clusters of user behaviour emerged: Fully Engaged, Consistent Viewers, One-Week Engaged, Two-Week Engaged, and Sporadic users. Assessment behaviours and scores are then examined within each cluster, and found different between clusters. Nonparametric statistical test, Kruskal-Wallis yielded a significant difference between user behaviour in each cluster. To make accurate inferences about what occurs in a MOOC, a first step is to understand the patterns of user behaviour. The latent characteristics that contribute to user behaviour must be explored in future research. Keywords: MOOCs, Learning Analytics, Assessment

AN EFFICIENT AND PRACTICALLY ROBUST HYBRID METAHEURISTIC ALGORITHM FOR SOLVING FUZZY BUS TERMINAL LOCATION PROBLEMS

Bus network design is an important problem in public transportation. In practice, some parameters of this problem are uncertain. We propose two models for the bus terminal location problem with fuzzy parameters. In the first formulation, the number of passengers corresponding to each node is a fuzzy number. In the second formulation, an additional assumption of fuzzy neighborhood is considered. These problems being NP-hard, we use a genetic algorithm (GA) and a simulated annealing (SA) algorithm for solving them. We also propose an idea to hybridize these algorithms. In our hybrid algorithm, SA is applied as a neighborhood search procedure of GA on the best individual of the population, which is the best available approximation of the optimal solution, with a varying probability that is gradually increased with the increase in the number of iterations in GA. We then implement GA, SA, our hybrid algorithm, and a recently proposed hybrid algorithm making use of a constant probability for application of SA on all the individuals of the population of GA, and use a nonparametric statistical test to compare their performances on a collection of randomly generated medium to large-scale test problems. Results of computational experiments demonstrating the efficiency and practicability of our proposed algorithm are reported.

A Pilot Study for the Statistical Treatment of Nutritional Parameters in the Quality Control of Canned Fishing Goods

This pilot study was derived as a consequence of European Directives 496/90 and 493/91 in which a regulation on the labeling of canned fishing goods was established. The study was intended primarily to assess whether different Spanish canned fishing goods might be differentiated by their basic nutritional composition (i.e., ash, chlorine as NaCl, fat, humidity, total proteins, and dry residue) and, second, to study each particular type of good. Accordingly, a univariate nonparametric statistical test and 2 multivariate chemometric techniques (factor and cluster analyses) were used. The pilot study revealed that (1) the basic nutritional variables did not allow a clear distinction among canned goods when different commodities were considered, but they seemed useful for obtaining information for only one type of good; and (2) the variables that gave the most useful information to visualize the appearance of groups in the data sets were humidity, dry residue, fat, and proteins, although their particular usefulness was found to be different when different species were considered.