Miguel Jiménez-Carrión

In the context of urban logistics, travel time uncertainty is a critical challenge for efficient route planning. Thus, this study proposes an algorithm based on Simulated Annealing (SA) that addresses this problem through a dual approach. On the one hand, the effectiveness of the algorithm is validated in deterministic VRPTW scenarios, using classical Solomon instances and achieving an average GAP of 3.42% in the most complex cases. On the other hand, a stochastic model fed with empirical data from Google Maps is integrated, designed to capture real-time traffic variability, thus addressing the VRPSTTW problem. The results show that the algorithm not only maintains a standard deviation of 2,048 energy units, consolidating its robustness to fluctuations in the optimal parameters, but also stands out for its ability to generate robust solutions in urban contexts with high temporal uncertainty. This proposal, being based on real data and not on theoretical simulations, positions the algorithm as a strategic tool to optimize logistic operations in dynamic and volatile environments.

This study aims to develop a hybrid algorithm using the ARIMA model and LSTM-type recurrent neural networks to predict the closing prices of the cryptocurrencies BTC, LTC, and ETH. The methodology includes an exploratory data analysis, followed by the design, implementation, and evaluation of each individual algorithm as well as the combined hybrid algorithm. The results, after experimentation and evaluation of metrics on the test set, indicated that the ARIMA model was inefficient in predicting the closing prices of cryptocurrencies. On the other hand, the hybrid model for BTC showed significant statistical differences in the metrics, with MAE = $726.21 and MAPE = 1.75%, compared to the LSTM model, which achieved MAE = $729.35 and MAPE = 1.76%. These results indicate better performance from the hybrid model. Regarding the RMSE metric, the hybrid model scored 1157.47, while LSTM scored 1159.99; although statistically equivalent, the hybrid model was numerically better. For the remaining metrics and other cryptocurrencies, both methods were statistically equivalent. For five-day-ahead predictions, the hybrid algorithm continued to yield better results for LTC and ETH.

This study aims to develop a predictive model for cryptocurrency prices in highly volatile markets. The methodology includes an exploratory data analysis, followed by designing and implementing machine learning (ML) algorithms, focusing on the Long Short-Term Memory (LSTM) neural network. The model's performance was optimized through hyperparameter tuning, and its stability was validated using an analysis of variance (ANOVA). We conducted a benchmark comparison with other ML approaches. Our LSTM model achieved an R² of 99.41% on the first day of prediction and maintained an accuracy above 97% up to the seventh day, demonstrating its robustness even for extended forecasts. During training, the LSTM model reached an RMSE of $1,187.14 and a MAPE of 2.20%, with the MAPE consistently remaining below 10% during the validation phase. For seven-day forecasts, the model recorded an RMSE of $5,038.46 and a MAPE of 6.83%. In comparison, alternative models such as Support Vector Machines (SVM), Extreme Gradient Boosting (XGBoost), and Random Forests exhibited significantly higher error rates; for instance, XGBoost recorded an RMSE of $17,849.66 and a MAPE of 27.74%. Overall, these findings highlight the superior performance of the LSTM model in addressing the challenges of cryptocurrency price forecasting. Doi: 10.28991/HIJ-2025-06-01-017 Full Text: PDF

The main objective of this study was to implement a computational prototype in two stages: the first stage primarily focused on generating efficient routes based on an evolutionary algorithm. In other words, the complex computational problem was solved in the first stage. The second stage then shifted its focus towards determining the fleet size and frequencies using an allocation algorithm. This approach was designed to address the complex combinatorial search problem within a public transportation network. In the first stage, the prototype utilizes the metaheuristic known as Genetic Algorithms (GA). Within the GA operators, an innovative method called "aggregated crossover" is employed, with an additional mutation procedure that maintains feasible descendants. In the second stage, an allocation algorithm is used, taking into account the routes generated in the first stage. The results demonstrate that in the first stage, the GA metaheuristic consistently delivers highly efficient routes in each run, confirming that the combinatorial complexity of the problem is effectively resolved in this initial phase. These results were validated on Mandl's Swiss Road network, showing superior solutions compared to those presented in previous studies. Notably, the execution time for this process is only 35 minutes.

The objective of this research was to construct a new generation syntheticregional development index (RDI) that encompasses the four interdependentcomponents of sustainable development (economic, social, environmental, andinstitutional). This index enables a comparative evaluation of regional development in Peru. The study employed rigorous methodologies, including Principal Component Analysis (PCA) for RDI construction, the Jenks Natural Breaksmethod for regional weighting, the sigma convergence analysis to examineregional disparities, and the Cronbach’s Alpha coefficient to assess reliability.The results reveal that for the 2015-2019 period, certain regions (Callao, Ica,Moquegua, Lima) exhibit triple the level of development compared to others(Cajamarca, Huancavelica, Puno, Loreto), despite the latter possessing substantial development potential. One contributing factor is the limited progress observed in the institutional dimension. Furthermore, in hierarchical order,the dimensions with the greatest contribution to sustainable development aresocial (95.20%), environmental (85.98%), economic (84.78%) and institutional(49.42%). The proposed solution includes a prototype web platform that utilizes intelligent algorithms, big data, web scraping, and geospatial information to iteratively showcase regional development in Peru.

El principal objetivo del presente trabajo fue implementar un prototipo computacional generador de rutas de transporte urbano, inspirado en la teoría de la evolución, para hacerle frente al problema de búsqueda combinatoria compleja de una red de transporte público. El prototipo utiliza el lenguaje de programación Python para la codificación y utiliza la metaheurística conocida como algoritmos genéticos (AG). Los resultados muestran que el prototipo computacional es preciso y rápido, entregando soluciones altamente eficientes en cada ejecución para el problema de transporte urbano. Los resultados fueron validados en la red Swiss Road de Mandl’s, demostrando ser mejores soluciones que los presentados en estudios anteriores. En conclusión, el proceso evolutivo del prototipo computacional converge soluciones altamente satisfactorias debido a los mecanismos innovadores de operadores genéticos de cruza, mutación y función de calidad, ponderando de igual forma a los parámetros que la integran.

This work introduces a fuzzy optimization model, which solves in an integrated way the berth allocation problem (BAP) and the quay crane allocation problem (QCAP). The problem is solved for multiple quays, considering vessels’ imprecise arrival times. The model optimizes the use of the quays. The BAP + QCAP, is a NP-hard (Non-deterministic polynomial-time hardness) combinatorial optimization problem, where the decision to assign available quays for each vessel adds more complexity. The imprecise vessel arrival times and the decision variables—berth and departure times—are represented by triangular fuzzy numbers. The model obtains a robust berthing plan that supports early and late arrivals and also assigns cranes to each berth vessel. The model was implemented in the CPLEX solver (IBM ILOG CPLEX Optimization Studio); obtaining in a short time an optimal solution for very small instances. For medium instances, an undefined behavior was found, where a solution (optimal or not) may be found. For large instances, no solutions were found during the assigned processing time (60 min). Although the model was applied for n = 2 quays, it can be adapted to “n” quays. For medium and large instances, the model must be solved with metaheuristics.

Se ha implementado un algoritmo genético simple para resolver el Problema de Programación de la Tienda de Trabajo (Job Shop Scheduling Problem, JSSP). El diseño del cromosoma representa una solución factible y cumple con todas las restricciones. Se utilizó el mecanismo de selección por torneo, una reproducción del 95% basada en emparejamiento parcial con dos puntos de cruza, una estrategia mixta en la etapa de mutación combinando el método de intercambio y el método de inversión usando dos puntos aleatorios en cada máquina y un porcentaje de mutación progresivo entre 2% al 5%. Los resultados muestran que el algoritmo debe ejecutarse con 100 individuos como tamaño de población y 500 generaciones para problemas cuyos tiempos de operación están entre 0 y 10 unidades de tiempo y con 100 individuos y 1500 generaciones para problemas entre 0 y 100 unidades de tiempo. El estudio muestra que el algoritmo implementado encuentra soluciones óptimas en el primer caso y soluciones altamente competitivas en el segundo caso. Estas son comparables con los resultados publicados en la literatura que por lo general son respuestas a algoritmos híbridos repotenciados con otras metaheurísticas.

Se ha aplicado redes neuronales artificiales a los datos climáticos de precipitación, temperaturas superficiales de mar en diferentes zonas calificadas como Niño y la velocidad de los vientos alisios con el fin de modelar y predecir el fenómeno El Niño, con seis meses de anticipación a la presencia del mismo. El estudio se realiza en Piura, Perú. Se realiza un análisis preliminar de la información para determinar el grado de correlación entre las variables, luego se ha diseñado el modelado en dos fases, la primera usa las redes neuronales para modelar las variables como series temporales y en la segunda fase se diseña una red neuronal usando MatLab para que simule el comportamiento de las precipitaciones en Piura. Los resultados del estudio muestran que las redes neuronales son una técnica altamente confiable para encontrar un patrón de comportamiento de las precipitaciones y luego para predecir el fenómeno alcanzando una tasa de acierto del 98.4% en la etapa de entrenamiento y de 100% en la predicción del primer semestre del año 2016.