Ciencias Exactas y Ciencias de la Salud

Permanent URI for this collectionhttps://hdl.handle.net/11285/551039

Pertenecen a esta colección Tesis y Trabajos de grado de las Maestrías correspondientes a las Escuelas de Ingeniería y Ciencias así como a Medicina y Ciencias de la Salud.

Browse

Search Results

Now showing 1 - 10 of 26

Adaptive learning for providing inclusive contents based on student profile in digital education
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12-03) Alvarado Reyes, Ignacio; Molina Espinosa, José Martín; emipsanchez; Icaza Longoria, Inés Alvarez; Suárez Brito, Paloma; School of Engineering and Sciences; Campus Estado de México
The application of artificial intelligence technologies in educational fields has been increasing in the last years, especially with the implementation of adaptive learning technologies, designed to monitor different characteristics of students and provide them with content and suggestions aimed at improving their performance and avoiding problems they may have on digital platforms. In this study, the reference framework for student classification was explored with a proposal of the contents and accessibility functions that could be applied based on their learning characteristics, complemented by an implementation of adaptive learning technologies consisting of a classifier based on the decision tree algorithm that automatically processes student data and classify them within the classes defined in the framework. For the implementation of the classifier, it was trained with two data sets, initially with data generated in the laboratory and later with experimental data, obtained through a survey aimed at higher education students. Both instances of the trained algorithm demonstrated high accuracy for the classification process (99.98% with synthetic data and 95.94% with experimental data). Subsequently, through the same survey, the suggestions related to the classes assigned to the students were validated, as well as the suggested accessibility features and content. The suggestions seem to have a favorable acceptance range with rejection percentages between 0% and 6% for the content selections and between 14% and 34% for the accessibility options. With this dynamic implementation of educational content and digital accessibility features, we seek to provide personalized learning for different student profiles while seeking to implement more features related to compliance with concerns about diversity and inclusion.
Enhancing BGP security with MAD anomaly detection system and machine learning techniques
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12) Romo Chavero, María Andrea; Cantoral Ceballos, José Antonio; emipsanchez; Botero Vega, Juan Felipe; Navarro Barrón, Francisco Javier; School of Engineering and Sciences; Campus Monterrey; Pérez Díaz, Jesús Arturo
Anomalies in the Border Gateway Protocol (BGP) represent a signicant vulnerability in the Internet’s infrastructure, as they can cause widespread disruptions, trafc misdirection, and even security breaches. Proactive detection of these anomalies is vital to preserving network stability and preventing potential cyberattacks. In response to this challenge, we present the Median Absolute Deviation (MAD) anomaly detection system, which combines traditional statistical methods with advanced machine learning (ML) techniques for more precise and dynamic detection. Our approach introduces a novel adaptive threshold mechanism, allowing the system to adjust based on the changing conditions of network trafc. This dynamic thresholding signif- icantly improves the accuracy, precision, and F1-score of anomaly detection compared to the previous xed-threshold version. Additionally, we integrate the MAD system with a diverse ML classiers, including Random Forest, XGBoost, LightGBM, CatBoost, and ExtraTrees to enhance the system’s ability to identify complex patterns that indicate unusual BGP behavior.We evaluate our detection system on well-documented BGP anomaly events, such as the Slammer worm, Nimda, Code Red 1 v2, the Moscow blackout, and the Telekom Malaysia misconguration. The results show that our system when combined with ML models achieves an overall accuracy and F1-score of 0.99, demonstrating its effectiveness across various anomaly types. By using both statistical and ML models, the system is able to capture irregularities that could signal security threats, offering a more comprehensive detection solution.This research highlights the importance of combining statistical anomaly detection with ML to obtain a balance between accuracy and computational efciency. The system’s low resource requirements and minimal pre-processing make it highly scalable, allowing it to be potentially deployed in real-time on large-scale networks.
Crowd-scouting: enhancing football talent identification through the use of machine learning and wisdom of crowds
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12) Díaz de León Rodríguez, Iván; Zareei, Mahdi; emimmayorquin; Roshan Biswal, Rajesh; School of Engineering and Sciences; Campus Estado de México; Hinojosa Cervantes, Salvador Miguel
The identification of talented young footballers is a cornerstone of success in professional football. This capability empowers established clubs to nurture potential superstars who elevate team performance and propel them towards championship contention. Smaller clubs strategically leverage this skill set to develop talent for an eventual sale, boosting their financial situation and, in some instances, even mounting their own title challenges. Ultimately, the ability to recognize future elite players has consistently translated into a significant competitive advantage throughout the history of the sport. This thesis delves into this domain by comparing the performance of three supervised machine learning models (Random Forest, Gradient Boosting, and Support Vector Machines). The models were trained using two comprehensive datasets encompassing data for 1,086 male professional footballers. The first one incorporates player statistics, game-related attributes, and transfer market values. The second one incorporates YouTube metrics to leverage the well-established concept of the wisdom of crowds. This concept presumes that the collective intelligence of a large group can outperform individual judgment. The wisdom of the fans has the potential to optimize scouting efforts. Historical and literary evidence suggests that the most effective strategies combine data with human judgment, particularly for complex tasks such as talent identification. SVM demonstrated the highest effectiveness, achieving superior sensitivity and identifying the greatest proportion of elite players within the dataset under the baseline scenario following a 5-fold cross-validation. Although its performance declined after the inclusion of crowd-sourced features, SVM continued to capture the largest portion of elite players, despite its lower precision score. The crowd-sourced features exhibited surprising potential when integrated with tree-based models, enhancing both sensitivity and precision in identifying the minority class. These models successfully captured a significantly larger share of the minority class while preserving their discriminative capacity. Integrating the collective knowledge of football fans improved the performance of a classification algorithm in identifying elite players using the selected features; thus, thereby validating the hypothesis stated in this dissertation. Furthermore, the feature importance analysis and other valuable insights gleaned from the study pave the way for further research endeavors. By providing this comparative analysis, the study aims to encourage the adoption of advanced data analytics, statistical methods, and more crowd-sourced data within football clubs worldwide. This approach can empower them to optimize resource allocation and refine their talent identification strategies.
Multimodal data fusion algorithm for image classification
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-11) Beder Sabag, Taleb; Vargas Rosales, César; emipsanchez; Pérez García, Benjamín de Jesús; School of Engineering and Sciences; Campus Monterrey
IImage classification algorithms are a tool that can be implemented on a variety of research sectors, some of these researches need an extensive amount of data for the model to obtain appropriate results. A work around this problem is to implement a multimodal data fusion algorithm, a model that utilizes data from different acquisition frameworks to complement for the missing data. In this paper, we discuss about the generation of a CNN model for image classification using transfer learning from three types of architectures in order to compare their results and use the best model, we also implement a Spatial Pyramid Pooling layer to be able to use images with varying dimensions. The model is then tested on three uni-modal data-sets to analyze its performance and tune the hyperparameters of the model according to the results. Then we use the optimized architecture and hyperparameters to train a model on a multimodal data-set. The aim of this thesis is to generate a multimodal image classification model that can be used by researchers and people that need to analyze images for their own cause, avoiding the need to implement a model for a specific study.
Machine translation for suicide detection: validating spanish datasetsusing machine and deep learning models
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-11) Arenas Enciso, Francisco Ariel; Zareel, Mahdi; emipsanchez; García Ceja, Enrique Alejandro; Roshan Biswal, Rajesh; School of Engineering and Sciences; Sede EGADE Monterrey
Suicide is a complex health concern that affects not only individuals but society as a whole. The application of traditional strategies to prevent, assess, and treat this condition has proven inefficient in a modern world in which interactions are mainly made online. Thus, in recent years, multidisciplinary efforts have explored how computational techniques could be applied to automatically detect individuals who desire to end their lives on textual input. Such methodologies rely on two main technical approaches: text-based classification and deep learning. Further, these methods rely on datasets labeled with relevant information, often sourced from clinically-curated social media posts or healthcare records, and more recently, public social media data has proven especially valuable for this purpose. Nonetheless, research focused on the application of computational algorithms for detecting suicide or its ideation is still an emerging field of study. In particular, investigations on this topic have recently considered specific factors, like language or socio-cultural contexts, that affect the causality, rationality, and intentionality of an individual’s manifestation, to improve the assessment made on textual data. Consequently, problems like the lack of data in non-Anglo-Saxon contexts capable of exploiting computational techniques for detecting suicidal ideation are still a pending endeavor. Thus, this thesis addresses the limited availability of suicide ideation datasets in non-Anglo-Saxon contexts, particularly for Spanish, despite its global significance as a widely spoken language. The research hypothesizes that Machine- Translated Spanish datasets can yield comparable results (within a ±5% performance range) to English datasets when training machine learning and deep learning models for suicide ideation detection. To test this, multiple machine translation models were evaluated, and the two most optimal models were selected to translate an English dataset of social media posts into Spanish. The English and translated Spanish datasets were then processed through a binary classification task using SVM, Logistic Regression, CNN, and LSTM models. Results demonstrated that the translated Spanish datasets achieved scores in performance metrics close to the original English set across all classifiers, with limited variations in accuracy, precision, recall, F1-score, ROC AUC, and MCC metrics remaining within the hypothesized ±5% range. For example, the SVM classifier on the translated Spanish sets achieved an accuracy of 90%, closely matching the 91% achieved on the original English set. These findings confirm that machine-translated datasets can serve as effective resources for training ML and DL models for suicide ideation detection in Spanish, thereby supporting the viability of extending suicide detection models to non-English-speaking populations. This contribution provides a methodological foundation for expanding suicide prevention tools to diverse linguistic and cultural contexts, potentially benefiting health organizations and academic institutions interested in psychological computation.
Emotion recognition based on physiological signals for Virtual Reality applications
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2022-06-13) Oceguera Cuevas, Daniela; FUENTES AGUILAR, RITA QUETZIQUEL; 229297; Fuentes Aguilar, Rita Quetziquel; puemcuervo; Antelis Ortíz, Javier Mauricio; Fernández Cervantes, Victor; School of Engineering and Sciences; Campus Monterrey; Hernández Melgarejo, Gustavo
Virtual Reality (VR) Systems have been used in the last years with an increasing frequency because they can be implemented for multiple applications in various fields. Some of these include aerospace, military, psychology, education, and entertainment. A way to increase the sense of presence is to induce emotions through the VE, and since one of the main purposes of VR Systems is to evoke the same emotions as a real experience would, the induction of emotions and emotion recognition could be used to enhance the experience. The emotion of a user can be recognized through the analysis and processing of physiological signals such as Electrocardiogram (ECG) and Electrodermal Activity (EDA) signals. However, very few systems that present online feedback regarding the subject’s emotional state and the possibility of adapting the VE during user experience have been developed. This thesis proposes the development of a Virtual Reality video game that can be dynamically modified according to the physiological signals of a user to regulate his emotional state. The first experiment served for the creation of a database. Previous studies have shown that specific features from these signals, can be used to develop algorithms capable of classifying the emotional states of the subjects into multiple classes or the two emotional dimensions: valence and arousal. Thus, this experiment helped to develop an appropriate Virtual Reality video game for stress induction, a signal acquisition, and conditioning system, a signal processing model and to extract time-domain signal features offline. A statistical analysis was performed to find significant differences between game stages and machine learning algorithms were trained and tested to perform classification offline. A second experiment was performed for the Proof of Concept Validation. For this, a model was created to extract features online and the classification algorithms were re-fitted with the online extracted features. Additionally, to facilitate a completely online process, the signal processing and feature extraction models were embedded on an STM32F446 Nucleo board, a strategy was implemented to dynamically modify the VE of the Virtual Reality video game according to the detected class, and the complete system was tested.
The identification of DoS and DDoS attacks to IoT devices in software defined networks by using machine learning and deep learning models
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2022-05) Almaraz Rivera, Josué Genaro; PEREZ DIAZ, JESUS ARTURO; 31169; Pérez Díaz, Jesús Arturo; puelquio/mscuervo; Trejo Rodríguez, Luis Ángel; Botero Vega, Juan Felipe; School of Engineering and Sciences; Campus Monterrey; Cantoral Ceballos, José Antonio
This thesis project explores and improves the current state of the art about detection techniques for Distributed Denial of Service (DDoS) attacks to Internet of Things (IoT) devices in Software Defined Networks (SDN), which as far as is known, is a big problem that network providers and data centers are still facing. Our planned solution for this problem started with the selection of strong Machine Learning (ML) and Deep Learning (DL) models from the current literature (such as Decision Trees and Recurrent Neural Networks), and their further evaluation under three feature sets from our balanced version of the Bot-IoT dataset, in order to evaluate the effects of different variables and avoid the dependencies produced by the Argus flow data generator. With this evaluation we achieved an average accuracy greater than 99% for binary and multiclass classifications, leveraging the categories and subcategories present in the Bot-IoT dataset, for the detection and identification of DDoS attacks based on Transport (UDP, TCP) and Application layer (HTTP) protocols. To extend the capacity of this Intrusion Detection System (IDS) we did a research stay in Colombia, with Universidad de Antioquia and in collaboration with Aligo (a cybersecurity company from Medellín). There, we created a new dataset based on real normal and attack traffic to physical IoT devices: the LATAM-DDoS-IoT dataset. We conducted binary and multiclass classifications with the DoS and the DDoS versions of this new dataset, getting an average accuracy of 99.967% and 98.872%, respectively. Then, we did two additional experiments combining our balanced version of the Bot-IoT dataset, applying transfer learning and a datasets concatenation, showing the differences between both domains and the generalization level we accomplished. Finally, we deployed our extended IDS (as a functional app built in Java and connected to an own cloud-hosted Python REST API) into a real-time SDN simulated environment, based on the Open Network Operating System (ONOS) controller and Mininet. We got a best accuracy of 94.608%, where 100% of the flows identified as attackers were correctly classified, and 91.406% of the attack flows were detected. This app can be further enhanced with the creation of an Intrusion Prevention System (IPS) as mitigation management strategy to stop the identified attackers.
Evolutionary clustering using classifiers: definition, implementation, scalability, and applications
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2022-04-22) Sainz Tinajero, Benjamin Mario; GUTIERREZ RODRIGUEZ, ANDRES EDUARDO; 436765; Guiterrez Rodriguez, Andres Eduardo; puemcuervo; Ortiz Bayliss, Jose Carlos; Amaya Contreras, Ivan Mauricio; Medina Perez, Miguel Angel; School of Engineering and Sciences; Campus Estado de México
Clustering is a Machine Learning tool for partitioning multi-dimensional data automatically into mutually exclusive groups, aiming to reflect the patterns of the phenomena it represents. Clustering algorithms perform this task conditioned by the clustering criterion modeled in its objective function. However, selecting the optimal criterion is a domain-dependent task that requires information on the cluster structure that a user often does not count on due to the unsupervised nature of the technique. Available approaches accentuate this problem as they perform clustering according to a similarity notion often limited to the concepts of compactness and connectedness, inducing bias and favoring clusters with certain shape, size, or density properties from using conventional distance functions. However, we cannot consider this a complete notion of a cluster because not every dataset will comply with both notions in the same proportion. Hence, research on this topic has not converged to a standard definition of a cluster, which raises the need for algorithms that produce adaptive solutions that mirror the underlying structures and relations within the data. This thesis is focused on the design of single-objective Evolutionary Clustering Algorithms that generate solutions that are not biased towards any cluster structure by optimizing a novel generalization clustering criterion. To achieve that, we designed objective functions modeled as a supervised learning problem, considering that a good partition should induce a well-trained classifier. That is how we decided to assess the quality of a clustering solution, according to its capability to train an ensemble of classifiers. The main contribution of this thesis is our series of Evolutionary Clustering Algorithms using Classifiers (the ECAC series), which introduces the aforementioned clustering criterion along with evolutionary computation. This meta-heuristic allows us to model distinct criteria to optimize while creating and evaluating multiple solutions along the process. The experimental results in the design of our family of methods ECAC, F1-ECAC, and ECAC-S, show an increase in similarity between the partitions created by our algorithms and the ground truth labels (obtained from the publicly available repositories where we retrieved the data) with a maximum Adjusted RAND Index of 0.96. Our second algorithm, F1-ECAC, proved the competitiveness of our contributions against traditional, single, and multi-objective Evolutionary Clustering algorithms showing no statistically significant difference against k-means, HG-means, and MOCLE. Our latest contribution, ECAC-S, was tested on a satellite image segmentation task, and it produced segmentations with higher average Adjusted RAND Index than k-means, Spectral-clustering, Birch, and DBSCAN in 4 out of 10 images.
Antimicrobial resistance prediction by bacterial genome-wide-association-study in non-fermenting bacilli with critical priority (Pseudomonas aeruginosa and acinetobacter baumannii).
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2021-12-07) Barlandas Quintana, Erick Alan; MARTINEZ LEDESMA, JUAN EMMANUEL; 200096; Martinez Ledesma, Juan Emmanuel; puemcuervo; González Mendoza, Miguel; Garza González, Elvira; School of Engineering and Sciences; Campus Estado de México; Cuevas Díaz Durán, Raquel
Antimicrobial resistance (AMR) (or drug resistance) is a natural phenomenon where microor- ganisms change their molecular, physical, or chemical structures to resist the drugs created by infections. The World Health Organization (WHO) had released for the first time a list of Multidrug-Resistant Bacteria (MRB) that pose the greatest threat to human health and for which new antibiotics are desperately needed. Acinetobacter baumannii and Pseudomonas aeruginosa resistant to carbapenems are part of the Gram-negative non-fermenting bacilli group with critical priority according to the WHO. For this, the final research purpose was to create and train a bioinformatic study capable of finding critical k-mers that could differentiate those strains of P. aeruginosa and A. baumannii resistant to carbapenems. Four k-mers sizes were performed for each bacterium (12, 14, 16, and 18), and two training and testing (70:30 and 80:20) schemas were used over seven different machine learning algorithms: Random Forrest, Adaboost, Xgboost, Decision Trees, Bagging Classifier, Support Vector Machine, and KNN. For both bacteria, the best models were obtained when using a k-mer length of 12. In the case of Acinetobacter baumannii, the best models obtained an accuracy of 0.99 for testing. Moreover, for Pseudomonas aeruginosa, the best accuracy obtained was 0.93 when us- ing Bagging Classifier. To investigate the sequences of the k-mers obtained, the National Cen- ter for Biotechnology Information (NCBI) Basic Local Alignment Search Tool BLAST was used. Ten to twenty sequences built with the k-mers were investigated for each model. When using a k-mer length of 12 for A. baumannii, 18 out of 20 sequences represented a crucial sequence in carbapenems (meropenem and imipenem) resistance. In the case of P. aerugi- nosa, 16 out of 20 sequences represented a key sequence. To complement this research, a Dynamic Programming algorithm was used to find changes over the reference genome that could explain the carbapenems resistance within the resistant genomes. Not all the resistant k-mer sequences were found over the reference genome, as some of them could be acquired by horizontal transference (Conjugation, Transformation, or Transduction inheritance). Fur- ther investigation over these sequences can be applied in creating new directed antibiotics or detecting easily resistant strains of Pseudomonas aeruginosa or Acinetobacter baumannii resistant to carbapenems.
Sweet pepper recognition and peduncle pose estimation
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2021-12-03) Montoya Cavero, Luis Enrique; Escobedo Cabello, Jesús Arturo; puemcuervo; Gómez Espinosa, Alfonso; School of Engineering and Sciences; Campus Monterrey; Díaz de León Torres, Rocío
As a result of the ongoing workforce decrease in the agricultural industry, there is an increased interest in harvesting robots for specialty crops. While a lot of research is available for produce detection, current vision systems still struggle to detect ripe produce under challenging conditions such as varying lighting and highly occluded scenarios. Because of this, research has mostly focused on improving the detection, localization, and orientation estimation by using state-of-the-art algorithms and sensors. In this document, a deep learning sweet pepper detection and pose estimation framework is proposed. The framework uses high-resolution colored images from an active RGB-D-based sensor to detect and segment individual green, red, orange, and yellow pepper, and their peduncle (produce stem) by pixel using a mask and region-based convolutional neural network. Then, using depth information from the sensor it estimates the pepper’s 3D location and the z-axis orientation of the camera reference frame if the peduncle is visible. Peduncle detection and localization are crucial as sweet pepper must be harvested individually by cutting the peduncle. Otherwise, if harvested by grasping (a grabbing and pulling motion) there is a high risk that the produce will be damaged. To validate the precision of the sweet pepper detection and pose estimation framework, a small-scale robotic arm was used. This process involved moving the end-effector towards either the peduncle, if it was detected, or the produce if no peduncle was detected. The deep learning framework presented here achieved competitive results compared to state-of-the-art sweet pepper vision subsystems: the whole recognition process (detection and localization) average time was of 0.86 seconds even though higher resolution images were used, the detection process obtained a 𝑚𝐴𝑃@50=0.602 (mean average precision with an intersection over union of 50 percent), and regarding the localization process, the vision subsystem obtained a pose estimation error of 𝑥:±28.75 𝑚𝑚,𝑦: ±21.25 𝑚𝑚,𝑎𝑛𝑑 ±15 𝑚𝑚 and regarding z-axis orientation of the camera reference frame an error of ±9.6°.

Ciencias Exactas y Ciencias de la Salud

Browse

Filters

Settings

Sort By

Results per page

Search Results