Search

The objective was to detect, model and predict the urban growth of population centers in different hydrological basins of Chihuahua, Mexico. Urban growth was modeled to predict 2030, 2040 and 2050 future scenarios using logistic regression and CA-Markov incorporated in the Land Change Modeler (LCM) integrated in IDRISI. Land use maps for the years 2010, 2015 and 2020 were produced using Support Vector Machine (SVM). The most intense land use changes (LULC) occurred in the period 2010-2015, the increase in urban areas comes mostly from forests. The RL model proved to be consistent showing McFadden's Pseudo R2 greater than 0.2. Receiver Operating Characteristic (ROC) values are considered acceptable. In the simulated urban land use, the highest growth is observed in the year 2030, and a low rate is identified for 2040 and 2050. [ABSTRACT FROM AUTHOR]

Colon cancer is a significant health issue in developed countries and ranks among the most common types of cancer. Early diagnosis of this disease increases the chances of survival for patients, while delayed diagnosis can lead to fatal outcomes. In this study, an EfficientNetB0 and Support Vector Machines (SVM) based model has been proposed for colon cancer detection. The EfficientNetB0 architecture is utilized to extract feature maps from histopathological images, and the SVM algorithm is employed to classify the obtained feature maps. Furthermore, to analyze the performance of the proposed model, a comparison is made with convolutional neural network (CNN) architectures such as EfficientNetB0, Xception, VGG19, InceptionV3, DenseNet121, and ResNet101. The datasets used for the study are the eight-class Kather-5k and the two-class LC25000 datasets. The findings indicate that the proposed model achieves higher success rates compared to existing CNN architectures on the Kather-5k dataset, with an accuracy of 99.70%, precision of 100%, recall of 100%, F1-Score of 100%, Gmean of 99.71%, specificity of 100%, and an AUC of 99.83%. Similarly, on the LC25000 dataset, the proposed model achieves 100% success rates in all metrics. When the combined dataset of Kather-5k and LC25000 is used, the proposed model demonstrates better performance compared to other models with an accuracy of 99.96%, precision of 100%, recall of 100%, F1-Score of 100%, G-mean of 99.92%, specificity of 100%, and an AUC of 99.96%. In addition, with the proposed model, a significant increase in success has been achieved in the success of the EfficientNetB0 architecture. [ABSTRACT FROM AUTHOR]

The use of quantitative techniques for the classification of population segments is a critical phase to evaluate their conditions. This information will serve as input for planning strategies to alleviate poverty. In this article, we present a model of vector support machines. Consequently, a sample of families residing in Cartagena de Indias is segmented, based on certain economic and sociodemographic variables. Analytical results confirm that most important factors are employment status, accessibility to public services and familiar income. In addition, it is corroborated that neighborhood conditions and monetary transfers have a low discriminatory power. [ABSTRACT FROM AUTHOR]

In this paper three prime numbers are presented as high potentials to be Mersenne numbers and their application in computational primality testing is suggested. These numbers are constructed from a regression algorithm based on Support vector machines (SVM) and using a Gaussian Kernel. Data training is carried out using the Phyton programming language, In the study we address the current data of Mersenne primes and work with the Ova-angular classification group for Mersenne primes 31. [ABSTRACT FROM AUTHOR]

Because of the impact on productivity and cost reduction, decision making in industrial processes is one of the most required aspects in the industry. Specifically in the panel industries, product quality depends on multiple variables, especially wood variability. Among other factors, quality depends on the adhesion of veneers or perpendicular tensile strength. The main objective of this study was to evaluate a Machine Learning approach to predict the adhesion under industrial conditions in the gluing and pre-pressing stage. The control variables that determine this adhesion are mainly: operational times, amount of adhesive, environmental conditions, and veneer temperature. Using Knowledge Discovery in Databases data analytics methodology, Artificial Neural Networks and Support Vector Machine were evaluated. Sigmoid activation function was used with 3 hidden layers and 245 neurons. In addition to the Adam optimizer, Multi-Layer Perceptron, Artificial Neural Networks delivered the best accuracy levels of over 66 %. Best result with Relu and Sigmoid functions were obtained. Sigmoid showed accuracy over 66 %, precision fit good to find positive results (70 %). Relu function obtained the best recall (over 74 %) showing a good capacity to identify reality. Results show that it is not sufficient to generate a data set using the averages of each process variable, since it is difficult to obtain better results with the algorithms evaluated. This work contributes to defining a methodology to be used in plywood plants using industrial data to train and validate Machine Learning models. [ABSTRACT FROM AUTHOR]

This study assesses spectral occupancy classification for decision-making processes by implementing feature extraction and classification rules. Cognitive radio (CR) is a technology that seeks to maximize the application of frequency resources by allowing unlicensed users to opportunistically access spectrum bands. The decisionmaking process is analyzed by classifying three traffic levels and by using three cost metrics and one benefit metric. The results show that the classifier using support vector machines presents the best performance, followed by KnNC (K-nearest Neighbor Classifier) and DAC (Discriminant Analysis Classifier). The worst performance, with the most deficient indicators' performance, is obtained by using BDT (Binary Decision Tree). It is concluded that CR offers a set of solutions that allows using the spectrum dynamically. [ABSTRACT FROM AUTHOR]

Credit score models quantify the risks in credit operations, customer segmentation, and approve or reject requests to credit customers. These models provide the necessary information to calculate the probabilities of default of any customer through the application of parametric or non-parametric techniques. This work identifies which model (Logit, Probit, Neural Networks, or Linear Support-Vector Machine (L-SVM)) may be more appropriate to measure the credit risk of individuals in a Family Benefit Fund located in Colombia. The results show Linear Support Vector Machine produces better performance, but Probit - Stepwise models are equally useful and they have the advantage of being interpreting the calibrated parameters. [ABSTRACT FROM AUTHOR]

Development of portable devices, that allows real-time detection of atrial fibrillation, requires the implementation of automatic pattern recognition algorithms and an appropriate methodology for their execution in embedded systems. In the present article, the performances of an artificial neural network, a machine vector support, a k-nearest neighbors algorithm and a hybrid classifier implemented on a single-board computer, were compared in terms of detection capacity of arrhythmia and time response associated with real-time execution. The MIT-BIH AFIB database was used to train and validate the algorithms. In advance, the extraction of parameters associated with the stationary wavelet transform was developed. Results between 92 % and 97 % for sensitivity and specificity, and time responses between 6 s and 7.1 s were found in this research. [ABSTRACT FROM AUTHOR]

The prediction of soil properties through spectral information is widely discussed in the current scientific literature. The objective of this review was to find algorithms with the highest predictive potential for soil physicochemical properties based on spectral information captured with different instruments. A systematic review was carried out in which 121 articles were found, and 19 of them were chosen which met a determination coefficient greater than 0.80 or a root mean square error close to 0. It was determined that the most used spectral range corresponds to the range from 350 to 2500 nm; the partial least squares, support vector machine, and adjusted support vector machine algorithms are suitable for predicting pH, organic matter, and organic carbon. Furthermore, linear regression is only effective in predicting calcium carbonate, organic matter, moisture, and water content using individual bands. [ABSTRACT FROM AUTHOR]

Classify human movement has become a technological necessity, where defining the position of a subject requires identifying the trajectory of the limbs and trunk of the body, having the ability to differentiate this position from other subjects or movements, which generates the need to have data and algorithms that help their classification. Therefore, the discriminant capacity of motion capture data in physical rehabilitation is evaluated, where the position of the subjects is acquired with the Microsoft Kinect and optical markers. Attributes of the movement generated with the Frenet Serret framework. Evaluating their discriminant capacity by means of support vector machines, neural networks, and k nearest neighbors algorithms. The obtained results present an accuracy of 93.5% in the classification with data obtained from the Kinect, and success of 100% for movements where the position is defined with optical markers. [ABSTRACT FROM AUTHOR]

Mobile application stores, such as Google Play Store, have a wide range of applications (apps) target to different customers. These digital platforms provide a rating mechanism for users to rate hosted apps and leave their reviews. User feedback contains valuable information that has an inevitable impact on the success of an application. Due to the large amount of data generated on this platform, natural language processing techniques have become applied more frequently. Sentiment analysis is a text classification task that can be used to analyze these reviews. Sentiment analysis has been conducted mainly for the English language, while in other languages like Spanish, there are just few attempts. In this work, we use a pre-trained BERT model to perform sentiment analysis for Spanish reviews on Google Play Store. Experiments using the BERT model show that after proper preprocessing steps it can achieve promising results in a Spanish dataset. An accuracy of 0.81 can be reached on average, even with limited data. [ABSTRACT FROM AUTHOR]

In the present article a hyperparameter optimization of a vectorial-support regression machine via adaptation of metaheuristics of a particle swarm is proposed. This method will be used so that a forecasting of the time series of the total amount of positive accumulated cases of COVID-19 in Bogotá, Colombia. In order to validate the performance of the method, a comparison with a regression vectorial-support machine whose hyperparameters have not been optimized will be made, being the metrics those of performance measurement like mean square error, mean absolute error, and determination coefficient. The proposed method finds itself at a greater level of performance when the mean square error value is that of 0,000045, the determination coefficient corresponds with the value of 0,998884 and the pvalue of 0,0015, for the nonparametric Wilcoxon test. Finally, applicability of these sorts of methods for forecasting of cases-behavior amidst epidemics is discussed. [ABSTRACT FROM AUTHOR]

Introduction-- Deforestation and disordered timber extraction endanger some vulnerable timber species. These prohibited species could be detected during their transportation process if surveillance and control entities had adequate monitoring instruments. Although methods for identifying wood species are reported in previous works, they are not applicable to sites far from the main cities. Objective-- In present work it is proposed to use electronic noses (chemical sensor arrays) in order to quickly identify wood species, from the volatile compounds their timbers emanate. Methodology-- The measurement of aromas is done by using an array of 16 chemical sensors, whose curves are the input to a feature estimation procedure. Then, principal component analysis is performed, to finally apply a classification strategy based on support vector machines. In contrast to previous works, in present work the samples collection conditions are closer to those found on real environments for which this work seeks to solve the problem. In addition, the number of samples is larger and more varied. However, the number of samples collected for each species is not balanced; thus, a data augmentation technique is applied to compensate the class imbalance. Results-- When carrying out the experiments, a performance of approximately 80% is found. Conclusions-- Although the promising results, greater efforts must be carried out in order to obtain a better performance. [ABSTRACT FROM AUTHOR]

Supreme audit institutions, and specifically its governing body, the International Organization of Supreme Audit Institutions (intosai), have promoted during the last four years a series of initiatives in the fiscal context aimed at the use of technologies and methods that are replicable and generate tangible results, thus reinforcing the surveillance and auditing processes carried out by supreme audit institutions. In this sense, the Comptroller General of the Republic of Colombia has been strengthening its technological infrastructure and technical capacities in order to improve and optimize its efforts in the monitoring of the resources of Colombian citizens. Although this task is not an easy one, this entity has managed to detect patterns of contractors who monopolize state contracting and are inserted into different economic sectors, without probably having the technical competence to fulfill stipulated contractual deeds. These subjects are known in the field of the General Comptroller's office as "multi-object" contractors. This article explains the construction of a data set of 1,998 records labeled by experts that correspond to education sector contracts. Training and tests were carried out with this tool on an automatic classifier built for the contractual objects in order to detect suspected "multi-object" contractors. It was found that the best classification algorithm was the "Linear Vector Support Machine," with an accuracy of 84%, which will eventually find presumed multi-object contractors by grouping. [ABSTRACT FROM AUTHOR]

Shrimp sector in Ecuador is nowadays one of the fastest-growing non-oil sectors towards the international market. In despite of this growth, to our knowledge most of the little producers of shrimps in Ecuador take important operational decisions based upon empirical knowledge, without considering historical data nor any scientific tool. In this work we implement and compare state-of-the-art statistical learning techniques for the prediction of shrimp harvest (in pounds) for a little shrimp farm located in Tenguel, Guayaquil-Ecuador. For this study we used historical information collected by the farm biologist. The data was organized and put into a digital format by the authors. Data from n=35 past harvests, corresponding to 7 cycles of production, were used to train the models. We then made predictions of shrimp harvest for the next two production cycles. We compare Multiple Linear Regression by means of ordinary least squares, CART Regression Tree, Random Forests, Multivariate Adaptive Regression Splines (MARS) and Support Vector Machines (SVM). In our analysis, MARS with no interaction terms allowed, Linear Regression with best subset variable selection and SVM with linear Kernel gave the lowest prediction error estimate by Cross Validation. Their good predictive performance was confirmed with good predictions on the next two production cycles. The use of statistical techniques can be of great help to improve predictions and therefore operational processes of small shrimp farms. [ABSTRACT FROM AUTHOR]

This paper presents the development of soft sensor empirical models using support vector machine (SVM) for the continual assessment of 2,3-dimethylbutane and 2-methylpentane mole percentage as important product quality indicators in the refinery isomerisation process. During the model development, critical steps were taken, including selection and pre-processing of the industrial process data, which are broadly discussed in this paper. The SVM model results were compared with dynamic linear output error model and nonlinear Hammerstein-Wiener model. Evaluation of the developed models on independent data sets showed their reliability in the assessment of the component contents. The soft sensors are to be embedded into the process control system, and serve primarily as a replacement during the process analysers' failure and service periods. [ABSTRACT FROM AUTHOR]

In this paper, new models for credit risk prediction in Colombia are proposed by using different artificial intelligence techniques. These models can be used to support the risk management area in banks, and they aim to identify clients that could be in default, generating a possible credit risk for financial institutions. Three techniques are used to obtain the models (neuronal networks, decision trees, and support vector machines) that predict the next payment of a client's fee based on basic data from the client and previous recorded installment payments. Decision trees turns out to be more accurate than the other techniques that have been used when predicting credit risk with a ROC area of 88.29%. The proposed models reach accuracies that are like some other papers in the state of the art and in some cases, they overcome models in other studies. [ABSTRACT FROM AUTHOR]

The purpose of this article is to make a brief introduction to five advanced machine learning prediction methods which may be useful for the development of prospective studies: logistic regression, support vector machines, gradient powered machines, random forests and neural networks. In addition, it is explained what methodology can be carried out to ensure robustness and validate these prediction models. As an example, it is presented how the use of these methods allowed to identify the most important financial variables to predict the development of innovation activities in Colombian SMEs. The results of the use of these methods may allow generating short and medium-term forecasts that serve to facilitate prospective studies with broader methods, such as the construction of scenarios, with the purpose of generating evidence-based proposals as a roadmap for long-term planning and public policy. [ABSTRACT FROM AUTHOR]

This article is derived from a research project in which an application for the prediction of ingredients and recipes by TensorFlow and support- vector machines was developed. A scheme with general architecture was developed, then a neural network was implemented, and then, the support-vector machine was run. After that, they were integrated via an application that allows the user to select ingredients' images for their prediction and the prediction of kitchens recipe in a didactic manner. It was concluded that the system has an average precision value of 75.8% and 71% for 17 ingredients categories and recipes classifier. In addition, sensitivity testing was performed on the application resulting on statistically equivalent results. [ABSTRACT FROM AUTHOR]

This article presents the design and development of a portable prototype for the acquisition, processing and classification of EEG signals with the aim of characterizing visual and auditory stimuli. Two different patients were worked with to validate the results, and the signals were recorded for 4 seconds at a frequency of 500Hz. The patients were exposed to visual and auditory stimuli in different cases, whose frequency of appearance remained constant. For the recording of the signals, a 4-channel acquisition system was designed, which could be configured to work with unipolar or bipolar derivation, as required by the experiment. The selection of the best base in the multi-resolution wavelet analysis, two important parameters were taken into account, the measurement of entropy and the percentages of classification of these levels, because the evoked potentials are generally constant in their morphology, it was made coherent averaging giving as a result the space-time location where this evoked potential appears, Once the characteristics of the treated signal were obtained, they were classified using two different methods of artificial intelligence, neural networks and vector support machines. At this stage, the measurement of the standard deviation of the data was taken into account to ensure that the learning machine was trained correctly. The results obtained reliably demonstrate the general behaviour of the evoked potentials as a result of the stimuli presented. In addition, it was possible to verify the variation of the patient's alpha waves according to his or her state of relaxation or alert in each case, it is advisable to carry out a much more robust filtering system to increase the signal-to-noise ratio of the EEG signal, facilitate its analysis and improve the results. [ABSTRACT FROM AUTHOR]

In this paper an explanation of the structure and how a self-organized neuro-fuzzy inference system (SONFIS) works, is given with detail. The study uses three classification problems (Fisher iris, Breast Cancer and Human Activities) to then compare the results with well-known universal classifiers such as artificial neural networks (ANN) and multiclass support vector machines (SVM). A brief description of each of these methods is presented. The results show that SONFIS has a similar, and sometimes better, performance than ANN and SVM with the advantage of generating a rule basis that helps understanding the inner structure of the problem. [ABSTRACT FROM AUTHOR]

The Colombian pension model is characterized by being intergenerational, reason why it has been affected by three major structural problems: inequality, low coverage and financial unsustainability, to which the Colombian government has reacted with regulatory measures that have somewhat relieved these problems. However, these prudential regulations have a negative effect on the inherent purpose of the investment fund, which is the generation of wealth to guarantee the intertemporal consumption of the Colombian elderly population. Therefore, the pension funds in Colombia requires that under the framework of the stablished regulations, new alternatives for structuring their portfolios be studied that allow them into better and more efficient management of retirement resources for Colombian people. In this sense, an efficient portfolio and frontier were built through a Support Vector Machine and the Black Litterman model to the Colombian Mandatory Pension Funds, with the aim to test the applicability of machine learning in financial fields. Results show that its applicability to the Black-Litterman portfolio structuring model is suitable through the improvement to the a priori distribution matrix, specifically with the use of Support Vector Regression, the model generated better diversified portfolios in comparison to the Markowitz model and shows to be adaptable to the Colombian Mandatory Pension Funds. [ABSTRACT FROM AUTHOR]

The theory of support vector machines applied to the classification of credits granted by the United States fintech banking LendingClub is presented. The classification given by the svm method to credits will be analyzed against what has already been granted by the entity. The most important variables that LendingClub manages for the granting of credits are analyzed, comparing it with the classification of payment or non-payment that the svm experiment gives. [ABSTRACT FROM AUTHOR]

The media coverage of armed conflicts tends to focus on the attribution of responsibilities and the humanitarian or security frames, as the main studies about media and conflicts have asserted. The present research analyzes the frames used in media outlets such as El Mundo and El País (Spain), HispanTV (Iran) and RT (Russia), in their coverage of wars in Yemen and Ukraine. Through innovative and machine learning techniques (particularly, SVM), the authors demonstrate that RT and HispanTV do not follow the traditional routines of mass media, and that the use of frames rely more on the geostrategic position of their countries of origin. [ABSTRACT FROM AUTHOR]

This research aims to identify changes in the writing style over time of 7 authors of Englishspeaking novels. For each author, an organization of the novels was carried out according to the date of publication. The novels were classified in three stages called initial, intermediate and final; each stage contains 3 novels. Between two consecutive stages there are at least 2 years of separation between the publication dates of the novels. To solve the problem of detecting changes in writing style over time, it is proposed to use a supervised automatic learning-based approach. Vector space models were created from the frequencies of use of n-grams of different types and lengths. In addition, the algorithm of Principal Component Analysis (PCA) was used as the n-gram selection method. The solution was addressed as a classification problem using the Vector Support Machine algorithms (Support Vector Machine, SVM), Naive Bayes Multinomial (Multinomial Naive Bayes, MNB), Logistic Regression (LG) and Liblinear as classifiers. The metric to measure the efficiency of the learning algorithms was accuracy. The research showed significant changes in five of the authors with an average accuracy between 70% and 80% in the different types of n-grams. [ABSTRACT FROM AUTHOR]

This document presents the evaluation of a method of classifying faults in finished products using the combination of color, shape and texture descriptors. Subsequently, a Multiclass Support Vector Machine (SVM-Support Vector Machine) is used to detect possible faults. To validate the model an annotated database is built capturing plastic bottles with 11 manufacturing situations, including bottles in good condition and bottles with imperfections such as tears, bumps, cracks, etc. A cross validation was used applying the Monte Carlo method in order to obtain the statistical relevance of the proposed method. The SVM configuration uses the "One-vs-All" multiclass methodology with Gaussian Radial Kernel. A comparison is made with other art state methods in order to show advantages and disadvantages of the proposal. This work allows us to see the contribution of each descriptor modality in the classification of faults, where an efficiency greater than 85% is observed, due to the fusion of the descriptors. [ABSTRACT FROM AUTHOR]

This paper presents a detection system which generate an alarm if exist an anomaly state in the stability control system of an aircraft. A servo type controller with integrator for the regulation of the Roll, Pitch and Yaw was designed starting from the lateral longitudinal and dimensional model. The models are excited with steps of random sizes and disturbed with different random signals to generate the database of suitable and non-suitable states. Radial base classifiers were trained and validated for its use in the detection of anomalous behaviors that may occur in the aircraft stability. The accuracy obtained with the classifiers was greater than the 93.33% for all the variables studied, indicating that classification techniques used, offer reliability in the determination of anomalous States in the simulation of the flight of the aircraft and that could be used in reall flights [ABSTRACT FROM AUTHOR]

Current social networks provide information with high correlation with events that are occurring worldwide. Twitter is a microblogging network of real time messages in which people post about various classes of events. A relevant topic is traffic congestion; user-generated content is useful to assist drivers in avoiding crowded areas. This work proposes a model to predict traffic-related events, based on a set of machine learning methods, in which a spatio-temporal dataset is obtained from Twitter messages. The training stage uses geocoded traffic events, in order to generate possible sites with traffic congestion at a given time. As a case study, partial areas of the Mexico City were taking into consideration. [ABSTRACT FROM AUTHOR]

Environmental noise is a big problem related to the environmental pollution in cities, which affects the quality of people life. In this paper, a methodology that uses an approach based on Volunteered Geographic Information (VGI) for the monitoring, analysis and prediction of environmental noise is proposed. It can be very useful to propose alternatives and initiatives that improve the life in a city. So, this work is composed of the following stages: data acquisition, analysis and, data processing, as well as the information visualization, considering the temporality of the same and taking into account macro and micro levels of analysis for the study surface. In addition, some details of the design and development of a geographic information system are presented, consisting of a web-mapping system, an application for mobile devices called "NoiseMonitor", geospatial analysis and machine learning methods (support vector machines and artificial neural networks) for the prediction of environmental noise; by using contextual information; that is, some data related to the city. This kind of work seeks to take advantage of the willingness of citizens to participate collaboratively to sense their environment and be considered as human sensors, which unlike traditional approaches, the cost associated with the development and implementation of this project is much lower. Likewise, a case study based on the Mexico City is presented and discussed, particularly the fourth quadrant of the Historic Center of the City, which is very representative for the variety of environmental noise that is generated in that area. The application domain of this approach is oriented towards big data from a collaborative perspective, Internet of Things and smart cities. [ABSTRACT FROM AUTHOR]

The exact identification of Mesoscale Convective Systems (MCS) is not a simple task. In this work, Support Vector Machines (SVM), Decision Trees (DT) and Random Forests (RF) non-parametric machine learning algorithms (ML) were applied to detect MCS, from a series of GOES-13 weather satellite images acquired on April 03, 2013 every half hour from 11:45 to 22:15 hours, Coordinated Universal Time (UTC), covering the Colombian territory. The results obtained by these methods were compared with a traditional method referred to as brightness temperature (BT). Accuracy assessment was conducted using STEP (shape, theme, edge, position), a method that evaluates geometric and thematic similarity between objects, using as reference a dataset of high accuracy data extracted from images of precipitation Tropical Rainfall Measuring Mission (TRMM). The aim of this study was to determine whether using information from several spectral channels of weather images, rather than from a single infrared channel (IR) as traditional techniques do, allows accurate detection of MCS. Experimental results show that the Decision Tree (DT) and Random Forest (RF) algorithms performed better than the IR-TB algorithm to detect MCS, while that the results of SVM algorithm, suggest that it use may not be favorable for practical applications. Accuracy ranges at 95% confidence levels to DT, RF, TB and SVM algorithms, at evaluated time's instants [12:15,15:15,18:15,21:15] by similarity metric ([shape], [theme], [edge], [position]) in their respective order were: DT ([89%-98%], [75%- 94%], [68%-96%], [80%-98%]), RF ([87%-95%], [73%-89%], [67%-87%], [78%- 90%]), IR-TB ([83%-95%], [68%-87%], [60%-81%], [72%-91%]) y SVM ([86%- 93%], [72%-83%], [65%-78%], [76%-81%]). The decision criteria of the classification model yielded by DT could be replicated several times in different dates without performing visual interpretation in each image, being very useful for operational applications under the approach presented here. [ABSTRACT FROM AUTHOR]

Cervical cancer can be cured if detected and treated early, for which the Pap test has been fundamental. In this context, technological aids can reduce the subjective nature of the diagnosis, although there are still several issues to be solved. Here we address two of them: the identification of the cytoplasm and the nucleus in a cell image, and the determination of a set of relevant features for the detection of injured cells. In this paper we present two contributions. First, we propose an interactive segmentation method based on multispectral morphological processing in which most misleading imperfections are eliminated with a simple interaction of the analyst. Second, we made an analysis of the relevance of certain variables that characterize the relative sizes of nucleus and cytoplasm, their shapes, textures and roughness of the edges. The analysis is based on performance measures of a detector that uses feature extraction through principal component analysis (PCA) and separation of normal and injured cells by a support vector machine (SVM). We have found that minimal interaction with the physician allows for much more accurate and reliable segmentation than purely automatic methods. On the other hand, we have found that the most important characteristics for detection of injured cells are the relative sizes of the nucleus and the cytoplasm and their form, while other features, such as texture and roughness, are less relevant. [ABSTRACT FROM AUTHOR]

In this paper we present the results of implementing an algorithm based on Support Vector Machine (SVM) on the National Instruments platform Compaq Rio, abbreviated cRio for the detection of Polyethylene terephthalate (PET) containers leakage. To accomplish this, the acquisition and processing the differential pressure signal is performed and then using a decision model to determine whether the package leaks or not. There are many tools for digital signal processing, however in the SVM classification applications have proven to be a very good alternative. The SVM requires a training process, performed based on signals which are known class to which they belong. After this training, a classification is performed in a new bottle. The final algorithm is implemented on the cRio National Instruments system for online application. [ABSTRACT FROM AUTHOR]

Prediction of exchange rates movement is regarded as a challenging task of financial time series prediction. Given the complexity involved in the dynamics of these markets, the implementation of efficient predictive models from the observed exchange rates data, is a difficult job. This study attempted to develop two efficient models of machine learning and compared their performances in predicting the direction of movement in the intraday US dollar - Colombian Peso exchange rate (USD/COP). The models are based on two classification techniques, artificial neural networks (ANN) and support vector machines (SVM). Technical indicators, candlesticks and p≠rice returns were selected as inputs of the proposed models. Comprehensive parameter setting experiments for both models were performed to improve their prediction performances. Experimental results showed that average performance of ANN model (57.8%) and SVM model (55.6%) were found significant in predicting the USD/COP exchange rate. [ABSTRACT FROM AUTHOR]

Accurate identification of precipitating clouds is a challenging task. In the present work, Support Vector Machines, Decision Trees and Random Forests algorithms were applied to discriminate between precipitating clouds and non-precipitating clouds from a satellite weather image GOES- 13 covering the Colombian territory. The objective of this study was to evaluate the performance of machine learning (ML) algorithms for digital classification of cloud masses in terms of thematic accuracy classification using the conventional Mahalanobis algorithm as benchmark. Results show that ML algorithms provide more accurate classification of cloud masses than conventional algorithms. The best accuracy was obtained using Random Forests (RF), with an overall thematic accuracy of 97%. Furthermore, the classification obtained with the RF algorithm was compared pixel-to-pixel with NASA Tropical Rainfall Measurement Mission (TRMM) rainfall estimates, obtaining an overall accuracy of 94%. ML algorithms can therefore be used to improve current precipitating clouds identification methods. [ABSTRACT FROM AUTHOR]

In this paper, a comparative analysis on the use of four metaheuristics for obtaining an optimal adjustment of a fault locator based on support vector machines (SVM), is presented. This research is aimed to determine those techniques which help to obtain the best performance at the specific problem of fault location. The proposed fault locator is tested in the 34 nodes IEEE power distribution system where the average precision obtained considering the best alternatives is around 99%, using a database of 13824 registers from single phase, phase to phase, two phase to ground and three phase faults. The comparison of the parameterization alternatives shows how those metahueristics based on population have better performance that those based on trajectory, having a good performance in all of the tested situations. [ABSTRACT FROM AUTHOR]

The support vector machines (SVM) method was used to classify the anchovy (Engraulis ringens) and common sardine (Strangomera bentincki) species detected in south-central Chile by means of acoustic equipment. For this, descriptors of fish schools (morphology, bathymetry, energy, spatial position) extracted from ecograms were used. In order to obtain precise classifications using this methodology, it was necessary to optimize the parameters Gaussian-Kernel γ and penalty term C by analyzing the effect of the calibration on the confusion matrices resulting from the classification of the species under study. The SVM method correctly classified 95.3% of anchovy and sardine schools. The optimal parameters of the Gaussian-Kernel γ and penalty C obtained with the proposed methodology were γ = 450 and C = 0.95. These parameters have an important influence over the confusion matrix and the final classifications percentages, suggesting the development of experimental protocols for calibrating these parameters in future applications of this methodology. In all the confusion matrices, the common sardine showed the lowest classification error. The bottom depth was the descriptor that was most sensitive to the SVM, followed by school-shore distance. [ABSTRACT FROM AUTHOR]

Time series prediction is an important research problem due to its implications in engineering, economics, finance and social sciences. An important topic about this problematic is the development of new models and its comparison with previous approaches in terms of forecast accuracy. Recently, support vector machines (SVM) have been used for time series prediction, but the reported experiences are limited and there are some problems related to its specification. The aim of this paper is to propose a novel technique for estimating some constants of the SVM usually fixed empirically by the modeler. The proposed technique is used to estimate several SVM with the aim of forecast five benchmark time series; the obtained results are compared with the statistics reported in other papers. The proposed method allow us to obtain competitive SVM for the time series forecasted in comparison with the results obtained using other most traditional models. [ABSTRACT FROM AUTHOR]

South American leaf blight (SALB) and Anthracnose (ANT) are two diseases that affect rubber crops, causing significant losses in production. In this research, the identification of the leaf area affected by ANT and SALB in the phenological stages B, C and D of the leaves was carried out, using hyperspectral images of healthy leaves, with ANT and / or SALB. For this purpose, images of the leaves were taken using Surface Optics' SOC 710-VP hyperspectral camera that provides 128 bands in the 400-1000 nm range. The free software MultiSpec was used to segment the areas of affectation by SALB and / or ANT and obtain samples of leaves images that allowed to train classification algorithms with machine learning, in this way, automatically identify the healthy areas, with SALB or ANT in the rubber leaves. Five classification algorithms were used: Random Forest (RF), Multilayer Perceptron (MLP), Support Vector Machines (SVM), K-Nearest Neighbors (KNN) and AdaBoost (AB). All machine learning and image processing algorithms were implemented in Python 3.7.4. With all the classifiers, the discrimination of healthy plant was achieved, with SALB and ANT for the phenological stage D; however, with RF the best results were obtained for the three phenological stages, so this algorithm was used to classify all the pixels of the leaves. This study showed that it is possible to identify SALB or ANT symptoms in rubber leaves using hyperspectral imaging and classification algorithms. [ABSTRACT FROM AUTHOR]