Oscar Benito-Pacheco
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Oscar Benito-Pacheco has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
- Hugo Vega-Huerta
Universidad Nacional Mayor de San Marcos
13 shared publications - Rubén Gil-Calvo
Faculty of Systems Engineering, Universidad Nacional Mayor de San Marcos, Lima 15081, Peru
6 shared publications - Antonio Mambret Luna Figueroa
Faculty of Systems Engineering, Universidad Nacional Mayor de San Marcos, Lima 15081, Peru
5 shared publications - Jorge Pantoja-Collantes
Faculty of Systems Engineering, Universidad Nacional Mayor de San Marcos, Lima 15081, Peru
5 shared publications - Luis Guerra-Grados
Faculty of Systems Engineering, Universidad Nacional Mayor de San Marcos, Lima 15081, Peru
5 shared publications - Percy De-La-Cruz-VdV
Universidad Nacional Mayor de San Marcos
5 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (21) , 11063
Deep Learning Architecture Comparison for Pneumonia Detection in Pediatric Chest X-Rays: Ensemble Learning and Cross-Population External Evaluation
Two assumptions dominate deep learning research on pneumonia detection from chest radiographs: that newer architectures outperform older ones, and that intra-dataset evaluation suffices to establish clinical utility. This study examined both on the pediatric Chest X-Ray Images dataset. Nine architectures, spanning classical, efficient-scaling, and modern convolutional networks and Vision Transformers, plus a weighted soft-voting ensemble, were compared under a unified protocol. No post-2020 architecture provided a detectable advantage: InceptionV3 (2016) achieved the best accuracy in the main experiment (98.18%), the order of the top five models changed in an independent retraining, and Bonferroni-corrected McNemar tests found no significant differences among them. With components and weights selected on validation data, the ensemble reached 97.95% accuracy (ROC-AUC 0.997), versus 97.84% for its best component. An external evaluation on an adult RSNA subset, which combines age, institutional, acquisition, and label shifts, revealed markedly different behavior: external accuracy ranged from 71.30% to 86.20%, and VGG16, among the most accurate models internally, was the least robust. Calibration, ablation, quantitative Grad-CAM, and efficiency analyses complemented these findings. In this retrospective setting, architectural recency did not predict performance and intra-dataset accuracy did not anticipate cross-population robustness; external evaluation on independent cohorts remains necessary before clinical use.
- 9 November 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (21) , 11063
Open AccessDeep Learning Architecture Comparison for Pneumonia Detection in Pediatric Chest X-Rays: Ensemble Learning and Cross-Population External Evaluation
Two assumptions dominate deep learning research on pneumonia detection from chest radiographs: that newer architectures outperform older ones, and that intra-dataset evaluation suffices to establish clinical utility. This study examined both on the pediatric Chest X-Ray Images dataset. Nine architectures, spanning classical, efficient-scaling, and modern convolutional networks and Vision Transformers, plus a weighted soft-voting ensemble, were compared under a unified protocol. No post-2020 architecture provided a detectable advantage: InceptionV3 (2016) achieved the best accuracy in the main experiment (98.18%), the order of the top five models changed in an independent retraining, and Bonferroni-corrected McNemar tests found no significant differences among them. With components and weights selected on validation data, the ensemble reached 97.95% accuracy (ROC-AUC 0.997), versus 97.84% for its best component. An external evaluation on an adult RSNA subset, which combines age, institutional, acquisition, and label shifts, revealed markedly different behavior: external accuracy ranged from 71.30% to 86.20%, and VGG16, among the most accurate models internally, was the least robust. Calibration, ablation, quantitative Grad-CAM, and efficiency analyses complemented these findings. In this retrospective setting, architectural recency did not predict performance and intra-dataset accuracy did not anticipate cross-population robustness; external evaluation on independent cohorts remains necessary before clinical use.
- Open Access
- 19 October 2026
- Published by MDPI AG in Future Internet
- Vol. 18 (10) , 565
IXPs as the Foundation of Connectivity and Digital Development
Internet Exchange Points (IXPs) constitute critical infrastructure for reducing dependence on external transit and improving network performance; however, their deployment in rural areas of Latin America and the Caribbean (LAC) remains limited, and the adoption of routing security mechanisms is incomplete. This study evaluates three dimensions: the regional evolution of IXPs, observable routing-security evidence among entities associated with IXPs, and the operational performance of a pilot deployment in Junín, Peru. The methodology combined a bibliographic analysis of sources indexed in the ACM Digital Library, IEEE Xplore, and Scopus with data from PeeringDB, Regional Internet Registry (RIR) delegation files, BGP tables, Looking Glass servers, RPKI validation, and MANRS records. For the Junín pilot, a before/after comparison was conducted during March 2024. Between 2016 and 2021, the number of active IXPs in LAC increased from 68 to 98 sites (+44%), while 60 IXPs met the inclusion criteria for the routing-security analysis in the available January–March 2023 extraction records; of these, 30 (50%) showed public evidence of at least one routing-security mechanism among their associated entities. In Junín, the observed mean RTT changed from 137 to 6.9 ms for the Meta measurements and from 138 to 43.8 ms for the Google measurements; because different endpoints were measured before and after deployment, these values represent observed route-specific changes rather than strictly comparable reductions and they do not constitute an isolated causal estimate of the peering effect because the design did not include a control site and, in Meta’s case, the regional cache was part of the same intervention. Netflix was assessed using traceroute because an equivalent ICMP measurement could not be obtained and, therefore, its results were not combined with the ping-based ones. The urban–rural comparison of transit costs relies on approximate values for Lima and Junín compiled from secondary information and pilot documentation, whose original official source could not be traced; it is presented only as an illustrative order of magnitude, not as a verified official or regional billing measurement. Taken together, the results provide site-specific empirical evidence from a rural pilot and a regional framework for studying traffic localization, routing security, and connectivity gaps while avoiding generalizing the site’s results to rural LAC as a whole.
- 19 October 2026
- Published by MDPI AG in Future Internet
- Vol. 18 (10) , 565
Open AccessIXPs as the Foundation of Connectivity and Digital Development
Internet Exchange Points (IXPs) constitute critical infrastructure for reducing dependence on external transit and improving network performance; however, their deployment in rural areas of Latin America and the Caribbean (LAC) remains limited, and the adoption of routing security mechanisms is incomplete. This study evaluates three dimensions: the regional evolution of IXPs, observable routing-security evidence among entities associated with IXPs, and the operational performance of a pilot deployment in Junín, Peru. The methodology combined a bibliographic analysis of sources indexed in the ACM Digital Library, IEEE Xplore, and Scopus with data from PeeringDB, Regional Internet Registry (RIR) delegation files, BGP tables, Looking Glass servers, RPKI validation, and MANRS records. For the Junín pilot, a before/after comparison was conducted during March 2024. Between 2016 and 2021, the number of active IXPs in LAC increased from 68 to 98 sites (+44%), while 60 IXPs met the inclusion criteria for the routing-security analysis in the available January–March 2023 extraction records; of these, 30 (50%) showed public evidence of at least one routing-security mechanism among their associated entities. In Junín, the observed mean RTT changed from 137 to 6.9 ms for the Meta measurements and from 138 to 43.8 ms for the Google measurements; because different endpoints were measured before and after deployment, these values represent observed route-specific changes rather than strictly comparable reductions and they do not constitute an isolated causal estimate of the peering effect because the design did not include a control site and, in Meta’s case, the regional cache was part of the same intervention. Netflix was assessed using traceroute because an equivalent ICMP measurement could not be obtained and, therefore, its results were not combined with the ping-based ones. The urban–rural comparison of transit costs relies on approximate values for Lima and Junín compiled from secondary information and pilot documentation, whose original official source could not be traced; it is presented only as an illustrative order of magnitude, not as a verified official or regional billing measurement. Taken together, the results provide site-specific empirical evidence from a rural pilot and a regional framework for studying traffic localization, routing security, and connectivity gaps while avoiding generalizing the site’s results to rural LAC as a whole.
- Open Access
- 12 October 2026
- Published by MDPI AG in Preprints.org
A Hybrid Agentic RAG Architecture with Knowledge Graphs and Text-to-SQL for Automating Citizen Services in a Public Scientific Institution
Large language models (LLMs) can produce plausible but incorrect content, a risk that is unacceptable when public institutions disclose scientific and legal information. This study presents and evaluates a hybrid agentic retrieval-augmented generation (RAG) system that automates first-level responses to citizen information requests at the Marine Institute of Peru (IMARPE). A baseline analysis of 196 requests showed a mean response time of 12 business days, above the 10-day legal limit, and identified 26.9% of them as routine, data-extractable queries. The system, developed under CRISP-DM, uses LangGraph to route each query to a LightRAG knowledge graph (Neo4j) built from institutional yearbooks, to a Text-to-SQL agent that queries the IMARSIS relational database through the Model Context Protocol (MCP), or to both, with Gemini 2.5 Flash as the foundation model and a referral protocol for out-of-scope requests. Five institutional domain experts rated 30 test cases, with mean scores of 4.45/5 for factual accuracy, 4.10/5 for completeness, and 4.57/5 for clarity. Mean end-to-end software latency was 20.43 s (maximum 49.20 s), and 30 pilot users rated usability at 80.42 on the System Usability Scale (Cronbach’s α = 0.96). These preliminary, single-site results indicate that combining graph-based and SQL-based grounding is feasible for routine citizen-information services; comparison with simpler baselines remains pending.
- 12 October 2026
- Published by MDPI AG in Preprints.org
Open AccessA Hybrid Agentic RAG Architecture with Knowledge Graphs and Text-to-SQL for Automating Citizen Services in a Public Scientific Institution
Large language models (LLMs) can produce plausible but incorrect content, a risk that is unacceptable when public institutions disclose scientific and legal information. This study presents and evaluates a hybrid agentic retrieval-augmented generation (RAG) system that automates first-level responses to citizen information requests at the Marine Institute of Peru (IMARPE). A baseline analysis of 196 requests showed a mean response time of 12 business days, above the 10-day legal limit, and identified 26.9% of them as routine, data-extractable queries. The system, developed under CRISP-DM, uses LangGraph to route each query to a LightRAG knowledge graph (Neo4j) built from institutional yearbooks, to a Text-to-SQL agent that queries the IMARSIS relational database through the Model Context Protocol (MCP), or to both, with Gemini 2.5 Flash as the foundation model and a referral protocol for out-of-scope requests. Five institutional domain experts rated 30 test cases, with mean scores of 4.45/5 for factual accuracy, 4.10/5 for completeness, and 4.57/5 for clarity. Mean end-to-end software latency was 20.43 s (maximum 49.20 s), and 30 pilot users rated usability at 80.42 on the System Usability Scale (Cronbach’s α = 0.96). These preliminary, single-site results indicate that combining graph-based and SQL-based grounding is feasible for routine citizen-information services; comparison with simpler baselines remains pending.
- Open Access
- 8 October 2026
- Published by MDPI AG in Preprints.org
Deep Learning Architecture Comparison for Pneumonia Detection in Pediatric Chest X-rays: Ensemble Learning and Cross-Population External Evaluation
Two assumptions dominate deep learning research on pneumonia detection from chest radiographs: that newer architectures outperform older ones, and that intra-dataset evaluation suffices to establish clinical utility. This study tested both. Nine architectures, spanning classical, efficient-scaling, and modern convolutional networks and Vision Transformers, plus a weighted soft-voting ensemble, were compared under a unified protocol on the pediatric Chest X-Ray Images dataset. None of the post-2020 architectures outperformed InceptionV3 (2016), which achieved the best individual accuracy (98.18%). The validation-weighted ensemble reached 98.52% accuracy and 99.22% pneumonia sensitivity on the internal test set; because its components were selected using test-set rankings, these values are upper-bound estimates. Bonferroni-corrected McNemar tests found no significant differences among the top-performing models. A cross-population external evaluation on an adult RSNA subset, which combines age, institutional, acquisition, and label shifts, revealed markedly different behavior: InceptionV3 retained 85.30% accuracy, whereas VGG16 fell to 59.60%, and the ensemble underperformed InceptionV3 alone. Ablation, Grad-CAM, and efficiency analyses complemented these findings. Architectural recency is not a proxy for performance, and external evaluation should be an architecture-selection criterion rather than an optional addendum.
- 8 October 2026
- Published by MDPI AG in Preprints.org
Open AccessDeep Learning Architecture Comparison for Pneumonia Detection in Pediatric Chest X-rays: Ensemble Learning and Cross-Population External Evaluation
Two assumptions dominate deep learning research on pneumonia detection from chest radiographs: that newer architectures outperform older ones, and that intra-dataset evaluation suffices to establish clinical utility. This study tested both. Nine architectures, spanning classical, efficient-scaling, and modern convolutional networks and Vision Transformers, plus a weighted soft-voting ensemble, were compared under a unified protocol on the pediatric Chest X-Ray Images dataset. None of the post-2020 architectures outperformed InceptionV3 (2016), which achieved the best individual accuracy (98.18%). The validation-weighted ensemble reached 98.52% accuracy and 99.22% pneumonia sensitivity on the internal test set; because its components were selected using test-set rankings, these values are upper-bound estimates. Bonferroni-corrected McNemar tests found no significant differences among the top-performing models. A cross-population external evaluation on an adult RSNA subset, which combines age, institutional, acquisition, and label shifts, revealed markedly different behavior: InceptionV3 retained 85.30% accuracy, whereas VGG16 fell to 59.60%, and the ensemble underperformed InceptionV3 alone. Ablation, Grad-CAM, and efficiency analyses complemented these findings. Architectural recency is not a proxy for performance, and external evaluation should be an architecture-selection criterion rather than an optional addendum.
- Open Access
- 26 May 2026
- Published by MDPI AG in Preprints.org
Bluetooth-Controlled Mobile Robot for Long-Distance Selfie Capture Using a Master-Slave Dual-Smartphone Architecture and Arduino
Conventional selfie sticks, typically limited to a maximum reach of 1.5 m, are unable to effectively capture large group photographs or wide-background scenes, frequently resulting in restricted framing and facial cropping. This work presents the design, implementation, and experimental validation of a Bluetooth-controlled mobile robot with five degrees of freedom capable of capturing photographs at distances of up to 10 m using a dual-smartphone master–slave architecture. The proposed system integrates an Arduino Nano microcontroller, an L298N motor driver, three SG90 servo motors, an HC-05 Bluetooth module, and a rechargeable Li-Ion battery mounted on a four-wheel chassis. A master smartphone remotely controls both the mobile platform and the articulated camera arm, while a slave smartphone mounted on the arm captures images through its high-resolution rear camera with real-time video streaming. Experimental evaluations demonstrated 100% command transmission success at distances up to 10 m, automatic emergency stopping within 100–110 ms after Bluetooth disconnection, accurate face centering with minimal servo adjustments, and approximately 45 min of battery autonomy. The proposed open-source, low-cost teleoperated platform extends the effective photographic range from 1.5 m to 10 m while providing potentially improved photographic quality due to rear-camera imaging characteristics and full remote framing control compared with conventional selfie accessories.
- 26 May 2026
- Published by MDPI AG in Preprints.org
Open AccessBluetooth-Controlled Mobile Robot for Long-Distance Selfie Capture Using a Master-Slave Dual-Smartphone Architecture and Arduino
Conventional selfie sticks, typically limited to a maximum reach of 1.5 m, are unable to effectively capture large group photographs or wide-background scenes, frequently resulting in restricted framing and facial cropping. This work presents the design, implementation, and experimental validation of a Bluetooth-controlled mobile robot with five degrees of freedom capable of capturing photographs at distances of up to 10 m using a dual-smartphone master–slave architecture. The proposed system integrates an Arduino Nano microcontroller, an L298N motor driver, three SG90 servo motors, an HC-05 Bluetooth module, and a rechargeable Li-Ion battery mounted on a four-wheel chassis. A master smartphone remotely controls both the mobile platform and the articulated camera arm, while a slave smartphone mounted on the arm captures images through its high-resolution rear camera with real-time video streaming. Experimental evaluations demonstrated 100% command transmission success at distances up to 10 m, automatic emergency stopping within 100–110 ms after Bluetooth disconnection, accurate face centering with minimal servo adjustments, and approximately 45 min of battery autonomy. The proposed open-source, low-cost teleoperated platform extends the effective photographic range from 1.5 m to 10 m while providing potentially improved photographic quality due to rear-camera imaging characteristics and full remote framing control compared with conventional selfie accessories.
- 20 April 2026
- Published by Springer Nature
- p. 331-345
Business Intelligence with Data Mart to Reduce Reporting and Indicator Generation Time
- 20 April 2026
- Published by Springer Nature
- p. 331-345
Business Intelligence with Data Mart to Reduce Reporting and Indicator Generation Time
- Open Access
- 19 September 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (18) , 10202
K-Nearest Neighbors Model to Optimize Data Classification According to the Water Quality Index of the Upper Basin of the City of Huarmey
Water quality in Peru is an increasing concern, particularly in the upper Huarmey watershed, which is affected by heavy metal contamination and untreated wastewater. This study proposes an automated classification approach using three supervised machine learning algorithms—K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Random Forest (RF)—to assess the water quality based on the Water Quality Index (WQI) of Peru. The experimental results show that KNN outperforms other methods, reaching an accuracy of 95.2%. The proposed system automates and improves the classification accuracy compared with manual methods based on Microsoft Excel. The methodology, performance metrics, dataset characteristics, and geographical context are detailed to ensure replicability. This algorithm assists decision-makers with environmental monitoring and public health protection.
- 19 September 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (18) , 10202
Open AccessK-Nearest Neighbors Model to Optimize Data Classification According to the Water Quality Index of the Upper Basin of the City of Huarmey
Water quality in Peru is an increasing concern, particularly in the upper Huarmey watershed, which is affected by heavy metal contamination and untreated wastewater. This study proposes an automated classification approach using three supervised machine learning algorithms—K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Random Forest (RF)—to assess the water quality based on the Water Quality Index (WQI) of Peru. The experimental results show that KNN outperforms other methods, reaching an accuracy of 95.2%. The proposed system automates and improves the classification accuracy compared with manual methods based on Microsoft Excel. The methodology, performance metrics, dataset characteristics, and geographical context are detailed to ensure replicability. This algorithm assists decision-makers with environmental monitoring and public health protection.
- 26 June 2025
- Published by Springer Nature
- p. 445-455
Detecting Dengue Infection’s Initial Stages Using Neural Networks
- 26 June 2025
- Published by Springer Nature
- p. 445-455
Detecting Dengue Infection’s Initial Stages Using Neural Networks
- Open Access
- 3 June 2025
- Published by International Association of Online Engineering (IAOE) in International Journal of Online and Biomedical Engineering (iJOE)
- Vol. 21 (07) , 138-151
Construction of a New Data Set of Pleural Fluid Cytological Images for Research
The limited availability of standardized datasets has hindered the implementation of artificial intelligence (AI) models in serous fluid cytology, particularly in pleural fluid analysis. In this paper, we present the construction of a dataset of pleural fluid cytology images. The objective is to generate a dataset of pleural fluid cytologic images validated by two pathologists and classified into five categories for cell diagnosis, which will be used to train AI models. As a methodology, the images represent pleural fluid cytology samples that have been prepared through medical procedures and transferred to slides, providing valuable information when evaluated under the microscope by medical specialists through cytological examination. We documented the entire process for building the pleural fluid cytological image dataset, from image capture, labeling, preprocessing, standardization, and uploading to public platforms. As a result, we obtained a pleural fluid cytology dataset based on the International System (TIS) criteria for reporting serous fluids, classifying samples into AUS, MAL, ND, NFM, and SFM. This dataset is intended to support medical research, deep learning applications in medical image analysis, and improved diagnostic methodologies.
- 3 June 2025
- Published by International Association of Online Engineering (IAOE) in International Journal of Online and Biomedical Engineering (iJOE)
- Vol. 21 (07) , 138-151
Open AccessConstruction of a New Data Set of Pleural Fluid Cytological Images for Research
The limited availability of standardized datasets has hindered the implementation of artificial intelligence (AI) models in serous fluid cytology, particularly in pleural fluid analysis. In this paper, we present the construction of a dataset of pleural fluid cytology images. The objective is to generate a dataset of pleural fluid cytologic images validated by two pathologists and classified into five categories for cell diagnosis, which will be used to train AI models. As a methodology, the images represent pleural fluid cytology samples that have been prepared through medical procedures and transferred to slides, providing valuable information when evaluated under the microscope by medical specialists through cytological examination. We documented the entire process for building the pleural fluid cytological image dataset, from image capture, labeling, preprocessing, standardization, and uploading to public platforms. As a result, we obtained a pleural fluid cytology dataset based on the International System (TIS) criteria for reporting serous fluids, classifying samples into AUS, MAL, ND, NFM, and SFM. This dataset is intended to support medical research, deep learning applications in medical image analysis, and improved diagnostic methodologies.
- Open Access
- 5 May 2025
- Published by MDPI AG in Preprints.org
K-Nearest Neighbors Model to Optimize Data Classification According to the Water Quality Index of the Upper Basin of the City of Huarmey
Water quality in the city of Huarmey-Lima-Peru is assessed through monitoring processes conducted by the National Water Authority (ANA). Traditional formula-based calculations and Excel macros, along with manual data validation, significantly extend the analysis and documentation time for evaluating water quality parameters. This study focuses on classifying water samples into Human Consumption (1-A2) and Animal Drinking (3-D2) categories by developing a machine learning model, specifically utilizing the K-Nearest Neighbors (KNN) algorithm. The primary objective is to enhance accuracy and efficiency in classifying water quality for human and animal consumption, aligning with Peru's Water Quality Index standards (WQI). The implementation of the KNN model demonstrated superior classification performance compared to traditional spreadsheet-based methods, achieving an accuracy of 75% with Excel-based classification and 90% using KNN. Additionally, the research evaluates the performance of KNN against Random Forest (RF) and Support Vector Machine (SVM). This approach significantly improved the speed and accuracy of water sample categorization, benefiting decision-making in water resource management for the Huarmey basin.
- 5 May 2025
- Published by MDPI AG in Preprints.org
Open AccessK-Nearest Neighbors Model to Optimize Data Classification According to the Water Quality Index of the Upper Basin of the City of Huarmey
Water quality in the city of Huarmey-Lima-Peru is assessed through monitoring processes conducted by the National Water Authority (ANA). Traditional formula-based calculations and Excel macros, along with manual data validation, significantly extend the analysis and documentation time for evaluating water quality parameters. This study focuses on classifying water samples into Human Consumption (1-A2) and Animal Drinking (3-D2) categories by developing a machine learning model, specifically utilizing the K-Nearest Neighbors (KNN) algorithm. The primary objective is to enhance accuracy and efficiency in classifying water quality for human and animal consumption, aligning with Peru's Water Quality Index standards (WQI). The implementation of the KNN model demonstrated superior classification performance compared to traditional spreadsheet-based methods, achieving an accuracy of 75% with Excel-based classification and 90% using KNN. Additionally, the research evaluates the performance of KNN against Random Forest (RF) and Support Vector Machine (SVM). This approach significantly improved the speed and accuracy of water sample categorization, benefiting decision-making in water resource management for the Huarmey basin.