Classification of Cucumber Leaves Based on Nitrogen Content Using the Hyperspectral Imaging Technique and Majority Voting.

Sajad Sabzi Razieh Pourdarbani Mohammad Hossein Rohban Alejandro Fuentes_Penna José Luis Hernández-Hernández Mario Hernández Hernández (2021, [Artículo])

Improper usage of nitrogen in cucumber cultivation causes nitrate accumulation in the fruit and results in food poisoning in humans; therefore, mandatory evaluation of food products becomes inevitable. Hyperspectral imaging has a very good ability to evaluate the quality of fruits and vegetables in a non-destructive manner. The goal of the present paper was to identify excess nitrogen in cucumber plants. To obtain a reliable result, the majority voting method was used, which takes into account the unanimity of five classifiers, namely, the hybrid artificial neural network¿imperialism competitive algorithm (ANN-ICA), the hybrid artificial neural network¿harmonic search (ANN-HS) algorithm, linear discrimination analysis (LDA), the radial basis function network (RBF), and the Knearest- neighborhood (KNN). The wavelengths of 723, 781, and 901 nm were determined as optimal wavelengths using the hybrid artificial neural network¿biogeography-based optimization (ANNBBO) algorithm, and the performance of classifiers was investigated using the optimal spectrum. The results of a t-test showed that there was no significant difference in the precision of the algorithm when using the optimal wavelengths and wavelengths of the whole range. The correct classification rate of the classifiers ANN-ICA, ANN-HS, LDA, RBF, and KNN were 96.14%, 96.11%, 95.73%, 64.03%, and 95.24%, respectively. The correct classification rate of majority voting (MV) was 95.55% for test data in 200 iterations, which indicates the system was successful in distinguishing nitrogen-rich leaves from leaves with a standard content of nitrogen.

artificial neural network cucumber hyperspectral imaging majority voting nitrogen INGENIERÍA Y TECNOLOGÍA CIENCIAS TECNOLÓGICAS TECNOLOGÍA DE LOS ALIMENTOS

Detección de eventos violentos en publicaciones de redes sociales

Detection of violent events in social media publications

Esteban Ponce León (2023, [Tesis de maestría])

En los últimos años, ha habido un interés creciente en el monitoreo de redes sociales para recopilar información y, en algunos casos, para examinar la ocurrencia de delitos. Sin embargo, gran parte de las investigaciones hasta ahora solo se han centrado en ciudades de EE. UU. o extranjeras, y por ende, en publicaciones y conjuntos de datos en inglés El objetivo principal de esta tesis es diseñar un método que permita la identificación de publicaciones de eventos violentos en español y en Twitter, utilizando información multimodal y técnicas de aumento de datos que mejoren el rendimiento de los modelos. Para esto, el trabajo de investigación se dividió en dos fases experimentales. La primera orientada a identificar publicaciones a partir de solo texto, explorando diferentes técnicas de aumento de datos para texto y modelos de aprendizaje máquina y profundo. En la segunda fase, se extendió el método propuesto para abordar la identificación en un contexto multimodal, es decir, considerando tanto los textos de los tweets como las imágenes compartidas que los acompañan. En este caso el método propuesto consideró utilizar descripciones textuales de las imágenes y abordar la problemática desde el dominio textual, además se hicieron 2 tipos de aumento de datos para cada tipo de información. La evaluación de los métodos se hizo utilizando las colecciones de la tarea de evaluación DA-VINCIS 2022 y 2023. Los resultados demostraron una mejora en el rendimiento de los modelos al considerar el uso de información multimodal y el uso de aumento de datos.

In recent years, there has been a growing interest in monitoring social networks to gather information and, in some cases, to examine the occurrence of crime. However, much of the research so far has only focused on US or foreign cities, and thus on English-language publications and data sets. The main objective of this thesis is to design a method that allows the identification of publications of violent events in Spanish and on Twitter, using multimodal information and data augmentation techniques that improve the performance of the models. For this, the research work was divided into two experimental phases. The first aimed at identifying publications from only text, exploring different data augmentation techniques for text and machine and deep learning models. In the second phase, the proposed method was extended to address identification in a multimodal context, that is, considering both the texts of the tweets and the shared images that accompany them. In this case, the proposed method considered using textual descriptions of the images and addressing the problem from the textual domain, in addition, 2 types of data augmentation were made for each type of information. The evaluation of the methods was done using the collections of the DA-VINCIS 2022 and 2023 evaluation task. The results demonstrated an improvement in the performance of the models when considering the use of multimodal information and the use of data augmentation.

Detección de Violencia, Redes Sociales, Aumento de Datos, Procesamiento del Lenguaje Natural, BERT, BETO, Descripción de Imágenes Violence Detection, Social Networks, Data Augmentation, Natural Language Processing, BERT, BETO, Image Captioning INGENIERÍA Y TECNOLOGÍA CIENCIAS TECNOLÓGICAS TECNOLOGÍA DE LOS ORDENADORES MODELOS CAUSALES MODELOS CAUSALES