Academic Research

Automatic Classification of Plant Electrophysiological Responses to Environmental Stimuli Using Machine Learning and Interval Arithmetic


Machine Learning has application in a wide number of areas. Some of these applications are causing a revolution in numerous fields of science. In this paper, we used ML classifiers to prove that is possible identify the plant stress before the physical signs such as leaf fall, reduced productivity and others. In order to verify this hypothesis, we used different types of electrical signals that can be very useful to identify environmental conditions. In this context, we showed that some specific frequencies can be observed on ideal-condition plants, but disappearing after stress-environments, such as cold, low-light and osmotic conditions. 

FEMaR: A Finite Element Machine for Regression Problems

Regression-based tasks have been the forerunner regarding the application of machine learning tools in the context of data mining. Problems related to price and stock prediction, selling estimation, and weather forecasting are commonly used as benchmarking for the comparison of regression techniques, just to name a few. In this work, we propose an efficient and effective regression technique based on the Finite Element Method (FEM) theory, hereinafter called Finite Element Machine for Regression (FEMaR). We show the proposed approach can obtain very competitive results when compared against some state-of-the-art regression techniques.

River Sediment Yield Classification Using Remote Sensing Imagery

The monitoring of water quality is essential to the mankind, since we strongly depend on such resource for living and working. The presence of sediments in rivers usually indicates changes in the land use, which can affect the quality of water and the lifetime of hydroelectric power plants. In countries like Brazil, where more than 70% of the energy comes from the water, it is crucial to keep monitoring the sediment yield in rivers and lakes. In this work, we evaluate some state-of- the-art supervised pattern recognition techniques to classify different levels of sediments in Brazilian rivers using satellite images.

Evaluation of Genotoxic Effects in Brazilian Agricultural Workers Exposed to Pesticides and Cigarette Smoke Using Machine-Learning Algorithms


A very common task in the science is hypothesis test, generally the researchers use statistical tools to prove a hypothesis, however in the last years the Machine Learning algorithms is being used  for this purpose. In this work we use ML classifiers  to evaluate genotoxic effects in rural workers who were exposed to cigarette smoke and/or pesticides to identify possible classification patterns in the exposure groups.  All of the algorithms displayed an excellent classification (accuracy > 80%). 

An Ensemble-Based Stacked Sequential Learning Algorithm for Remote Sensing Imagery Classification

Image classification is used in many areas. Extracting information from images is a very important step, however doing so is not so simple. For example, how can a red pixel that belongs to a apple be differentiated from a red pixel that belongs to a tomato? It is necessary to use contextual-based image classification that considers spatial/temporal information during the learning process in order to make the classification process smarter. Sequential learning techniques are one of the most used techniques to perform contextual classification, being based on a two-step classification process. 

Pruning Optimum-Path Forest Ensembles Using Metaheuristic Optimization for Land-Cover Classification

Machine learning techniques have been actively pursued in the last years, mainly due to the increasing number of applications that make use of some sort of intelligent mechanism for decision-making processes. In this context, we shall highlight ensemble pruning strategies, which provide heuristics to select from a collection of classifiers the ones that can really improve recognition rates and provide efficiency by reducing the ensemble size prior to combining the model. In this article, we presented and validated an ensemble pruning approach. 

Automatic Identification of Epileptic EEG Signals through Binary Magnetic Optimization Algorithms

Epilepsy is a class of chronic neurological disorders characterized by transient and unexpected electrical disturbances of the brain. The automated analysis of the electroencephalogram (EEG) signal can be instrumental for the proper diagnosis of this mental condition. In this context, we apply Machine Learning classifiers in well-known EEG benchmark dataset composed of five classes of EEG signals. The experimental results evidenced the robustness of the proposed methodology in automatic identification of epileptic.

Quaternion-Based Deep Belief Networks Fine-Tuning

Deep learning techniques have been paramount in the last years, mainly due to their outstanding results in a number of applications. In this paper, we address the issue of fine-tuning parameters of Deep Learning classifiers by means of meta-heuristics due the fact that those classifiers are very sensitive to small changes in the input parameters. Also, we proposed two novel approaches that outperform the state-of-the-art results.

Social-Spider Optimization-Based Support Vector Machines Applied for Energy Theft Detection

Some problems require moderns and complexes algorithms, in this context  we combined the optimization (fine-tuning) and feature selection to obtain an accurate method able to identify energy theft with mean accuracy of 99%.

Fine-Tuning Contextual-Based Optimum-Path Forest for Land-Cover Classification

Contextual-based learning aims at considering neighboring pixels to improve pixel wise-oriented classification techniques. In this work, we presented a new framework and a post-processing procedure to avoid overcorrection over specific regions. The proposed approach outperformed previous results obtained with a standard classifier in satellite imagery. 

A New Approach to Contextual Learning Using Interval Arithmetic and Its Applications for Land-Use Classification

Contextual-based classification has been paramount in the last years, since spatial and temporal information play an important role during the process of learning the behavior of the data. In this work, we introduced Interval Arithmetic to the context of land-use classification in satellite images by describing a given sample and its neighbors using an interval of values, thus allowing a better representation of the model. The experimental results showed that we can considerably improve sequential learning-oriented pattern classification using concepts from Interval Arithmetic.

A Hyperheuristic Approach for Unsupervised Land-Cover Classification

Unsupervised land-cover classification is of great interest, since it becomes even more difficult to obtain high-quality labeled data. Still considered one of the most used clustering techniques, the well-known k-means plays an important role in the pattern recognition community. It is simple formulation and good results in a number of applications have fostered the development of new variants and methodologies to address the problem of minimizing the distance from each dataset sample to its nearest centroid (mean). In this paper, we present a hyperheuristic approach to combine different techniques used to enhance k-means effectiveness. 

Unsupervised Non-Technical Losses Identification through Optimum-Path Forest

Non-technical losses (NTL) identification has been paramount in recent years mainly due energy companies wanting to reduce their costs.  In this paper, we apply a specific clustering algorithm to identify irregular and regular profiles of commercial and industrial consumers obtained from a Brazilian electrical power company. Additionally, a model for the problem of NTL recognition as an anomaly-detection task has been proposed, when there is little or no information available about irregular consumers. The experimental results have shown the robustness of methodology for both unsupervised NTL recognition and anomaly-detection problems.

Robust Automated Cardiac Arrhythmia Detection in ECG Beat Signals

Millions of people are affected by heart diseases worldwide, whereas a considerable amount of them could be aided through an electrocardiogram (ECG) trace analysis, which involves the study of arrhythmia impacts on electrocardiogram patterns. In this work, we carried out the task of automatic arrhythmia detection in ECG patterns by means of supervised machine learning techniques. The experimental results revealed a high skill on generalizing data and excellent accuracy.

A New Computer Vision-Based Approach to Aid the Diagnosis of Parkinson's Disease

Even today, pointing out an exam that can diagnose a patient with Parkinson's disease (PD) accurately enough is not an easy task. Although a number of techniques have been used in search for a more precise method, detecting such illness and measuring its level of severity early enough to postpone its side effects are not straightforward. In this work, after reviewing a considerable number of works, we conclude that only a few techniques address the problem of PD recognition by means of micrography using computer vision techniques. Therefore, we consider the problem of aiding automatic PD diagnosis by means of spirals and meanders filled out in forms. 

Machine Learning Algorithms for Automatic Classification of Marmoset Vocalizations

Machine Learning Algorithms for Automatic Classification of Marmoset Vocalizations .jpg
Automatic classification of vocalization type could potentially become a useful tool for acoustic the monitoring of captive colonies of highly vocal primates. However, for classification to be useful in practice, a reliable algorithm that can be successfully trained on small datasets is necessary. In this work, we consider seven different classification algorithms with the goal of finding a robust classifier that can be successfully trained on small datasets. We obtained good classification  performance (accuracy > 0.83) even with a very small dataset.