Comparación de los modelos LightGBM y XGBoost para la predicción a la resistencia a la compresión del concreto
DOI:
https://doi.org/10.53673/th.v5i1.427Palabras clave:
Machine Learning, Grdient-boosting, diseño de mezcla, modelos y Grid Search.Resumen
El uso de modelos de Machine Learning (ML), particularmente aquellos basados en gradient-boosting, está transformando la predicción de propiedades materiales con relaciones no lineales complejas. Este artículo presenta un análisis exhaustivo del desempeño de dichos modelos en la predicción de la resistencia a la compresión del concreto en función de las características del diseño de mezcla utilizando 525 datos experimentales obtenidos del registro histórico del Laboratorio N°1 de Ensayo de Materiales "Ing. Manuel Gonzales de la Cotera" de la Facultad de Ingeniería Civil de la Universidad Nacional de Ingeniería (LEM), correspondientes al período 2015-2022. La base de datos se utilizó para entrenar y probar dos modelos de ML optimizados los cuales fueron Light GBM y XGBoost. Los modelos se entrenaron utilizando la optimización de Grid Search para ajustar los hiperparámetros de configuración y obtener el mejor rendimiento de para ambos modelos. Los resultados muestran que el modelo XGBoost tiene mayor rendimiento tanto en la fase de entrenamiento como en la de prueba en comparación al modelo Light GBM en cuanto a los parámetros estadísticos como RMSE, MAE, MAPE y R2.
Citas
O. Alshboul, G. Almasabha, A. Shehadeh, y K. Al-Shboul, «A comparative study of LightGBM, XGBoost, and GEP models in shear strength management of SFRC-SBWS», Structures, vol. 61, p. 106009, mar. 2024, doi: 10.1016/j.istruc.2024.106009.
A. Khashman y P. Akpinar, «Non-Destructive Prediction of Concrete Compressive Strength Using Neural Networks», Procedia Computer Science, vol. 108, pp. 2358-2362, ene. 2017, doi: 10.1016/j.procs.2017.05.039.
K. C. Mouli et al., «Performance analysis of linear and non-linear machine learning models for forecasting compressive strength of concrete», Cogent Engineering, vol. 11, n.o 1, p. 2368101, dic. 2024, doi: 10.1080/23311916.2024.2368101.
S. Pandey, S. Paudel, K. Devkota, K. Kshetri, y P. G. Asteris, «Machine learning unveils the complex nonlinearity of concrete materials’ uniaxial compressive strength», International Journal of Construction Management, vol. 0, n.o 0, pp. 1-15, abr. 2024, doi: 10.1080/15623599.2024.2345008.
Y. Pawar y Shrikant Rajendra Kate, «Curing of Concrete: A Review», 2020, doi: 10.13140/RG.2.2.32095.07848.
V. Cominato, F. Benavente Canteras, L. Andréia Gachet, y R. Cristina Cecche Lintz, «The effect of granulometry of natural and recycled coarse aggregate on permeable concrete properties», Materials Today: Proceedings, vol. 65, pp. 1711-1718, ene. 2022, doi: 10.1016/j.matpr.2022.04.717.
F. Cao et al., «Compressive strength prediction of Portland cement clinker using machine learning by XRD data», Journal of Sustainable Cement-Based Materials, vol. 13, n.o 8, pp. 1120-1131, ago. 2024, doi: 10.1080/21650373.2024.2351461.
H. Constantinescu, O. Gherman, C. Negrutiu, y S. P. Ioan, «Mechanical Properties of Hardened High Strength Concrete», Procedia Technology, vol. 22, pp. 219-226, ene. 2016, doi: 10.1016/j.protcy.2016.01.047.
B. W. Chong et al., «Design of Experiment on Concrete Mechanical Properties Prediction: A Critical Review», Materials, vol. 14, n.o 8, Art. n.o 8, ene. 2021, doi: 10.3390/ma14081866.
F. Khademi, M. Akbari, S. M. Jamal, y M. Nikoo, «Multiple linear regression, artificial neural network, and fuzzy logic prediction of 28 days compressive strength of concrete», Front. Struct. Civ. Eng., vol. 11, n.o 1, pp. 90-99, mar. 2017, doi: 10.1007/s11709-016-0363-9.
J. L. Santamaría, B. Adame, C. Bermeo, J. L. Santamaría, B. Adame, y C. Bermeo, «Influencia de la calidad de los agregados y tipo de cemento en la resistencia a la compresión del hormigón dosificado al volumen», Revista Digital Novasinergia, vol. 4, n.o 1, pp. 91-101, may 2021, doi: 10.37135/ns.01.07.05.
M. Z. Naser, «Integrating Machine Learning Models into Building Codes and Standards: Establishing Equivalence through Engineering Intuition and Causal Logic», Journal of Structural Engineering, vol. 150, n.o 5, p. 04024039, may 2024, doi: 10.1061/JSENDH.STENG-12934.
A. H. A. Ahmed, W. Jin, y M. A. H. Ali, «Prediction of compressive strength of recycled concrete using gradient boosting models», Ain Shams Engineering Journal, vol. 15, n.o 9, p. 102975, sep. 2024, doi: 10.1016/j.asej.2024.102975.
P. C. Chiadighikaobi, M. Hematibahar, M. Kharun, N. A. Stashevskaya, y K. Camara, «Predicting mechanical properties of self-healing concrete with Trichoderma Reesei Fungus using machine learning», Cogent Engineering, vol. 11, n.o 1, p. 2307193, dic. 2024, doi: 10.1080/23311916.2024.2307193.
L. A. Bedriñana, J. Sucasaca, J. Tovar, y H. Burton, «Design-Oriented Machine-Learning Models for Predicting the Shear Strength of Prestressed Concrete Beams», Journal of Bridge Engineering, vol. 28, n.o 4, p. 04023009, abr. 2023, doi: 10.1061/JBENF2.BEENG-6013.
B. P. Koya, S. Aneja, R. Gupta, y C. Valeo, «Comparative analysis of different machine learning algorithms to predict mechanical properties of concrete», Mechanics of Advanced Materials and Structures, vol. 29, n.o 25, pp. 4032-4043, oct. 2022, doi: 10.1080/15376494.2021.1917021.
H. Wang, J. Tang, M. Wu, X. Wang, y T. Zhang, «Application of machine learning missing data imputation techniques in clinical decision making: taking the discharge assessment of patients with spontaneous supratentorial intracerebral hemorrhage as an example», BMC Med Inform Decis Mak, vol. 22, n.o 1, p. 13, dic. 2022, doi: 10.1186/s12911-022-01752-6.
P. Lu, S. Chen, y Y. Zheng, «Artificial Intelligence in Civil Engineering», Mathematical Problems in Engineering, vol. 2012, n.o 1, p. 145974, ene. 2012, doi: 10.1155/2012/145974.
T. Han, A. Siddique, K. Khayat, J. Huang, y A. Kumar, «An ensemble machine learning approach for prediction and optimization of modulus of elasticity of recycled aggregate concrete», Construction and Building Materials, vol. 244, p. 118271, may 2020, doi: 10.1016/j.conbuildmat.2020.118271.
S. Moon y A. Munira Chowdhury, «Utilization of Prior Information in Neural Network Training for Improving 28-Day Concrete Strength Prediction», Journal of Construction Engineering and Management, vol. 147, n.o 5, p. 04021028, may 2021, doi: 10.1061/(ASCE)CO.1943-7862.0002047.
M. Mohtasham Moein et al., «Predictive models for concrete properties using machine learning and deep learning approaches: A review», Journal of Building Engineering, vol. 63, p. 105444, ene. 2023, doi: 10.1016/j.jobe.2022.105444.
Q. Wang, J. Qi, S. Hosseini, H. Rasekh, y J. Huang, «ICA-LightGBM Algorithm for Predicting Compressive Strength of Geo-Polymer Concrete», Buildings, vol. 13, n.o 9, Art. n.o 9, sep. 2023, doi: 10.3390/buildings13092278.
H.-V. T. Mai, M. H. Nguyen, y H.-B. Ly, «Development of machine learning methods to predict the compressive strength of fiber-reinforced self-compacting concrete and sensitivity analysis», Construction and Building Materials, vol. 367, p. 130339, feb. 2023, doi: 10.1016/j.conbuildmat.2023.130339.
S. Li y J. Y. Richard Liew, «Experimental and Data-Driven analysis on compressive strength of steel fibre reinforced high strength concrete and mortar at elevated temperature», Construction and Building Materials, vol. 341, p. 127845, jul. 2022, doi: 10.1016/j.conbuildmat.2022.127845.
X.-Y. Zhao, J.-X. Chen, y B. Wu, «An interpretable ensemble-learning-based open source model for evaluating the fire resistance of concrete-filled steel tubular columns», Engineering Structures, vol. 270, p. 114886, nov. 2022, doi: 10.1016/j.engstruct.2022.114886.
Z. Mei, J. Yu, W. Ding, L. Kong, y J. Zhao, «Bus Passenger Volume Forecasting Model Based on XGBoost Integrated Learning Algorithm», pp. 3100-3113, dic. 2020, doi: 10.1061/9780784483053.261.
A. Abdulalim Alabdullah, M. Iqbal, M. Zahid, K. Khan, M. Nasir Amin, y F. E. Jalal, «Prediction of rapid chloride penetration resistance of metakaolin based high strength concrete using light GBM and XGBoost models by incorporating SHAP analysis», Construction and Building Materials, vol. 345, p. 128296, ago. 2022, doi: 10.1016/j.conbuildmat.2022.128296.
M. Nithurshan y Y. Elakneswaran, «A systematic review and assessment of concrete strength prediction models», Case Studies in Construction Materials, vol. 18, p. e01830, jul. 2023, doi: 10.1016/j.cscm.2023.e01830.
Q.-F. Li y Z.-M. Song, «High-performance concrete strength prediction based on ensemble learning», Construction and Building Materials, vol. 324, p. 126694, mar. 2022, doi: 10.1016/j.conbuildmat.2022.126694.
J. Pal y D. Chakrabarty, «Effects of input/output parameters on artificial neural network model efficiency for breakthrough contaminant prediction», Water Supply, vol. 21, n.o 7, pp. 3614-3628, abr. 2021, doi: 10.2166/ws.2021.125.
S. Paudel, A. Pudasaini, R. K. Shrestha, y E. Kharel, «Compressive strength of concrete material using machine learning techniques», Cleaner Engineering and Technology, vol. 15, p. 100661, ago. 2023, doi: 10.1016/j.clet.2023.100661.
A. Behnood y E. M. Golafshani, «Machine learning study of the mechanical properties of concretes containing waste foundry sand», Construction and Building Materials, vol. 243, p. 118152, may 2020, doi: 10.1016/j.conbuildmat.2020.118152.
C. Deepa, K. SathiyaKumari, y V. P. Sudha, «Prediction of the Compressive Strength of High Performance Concrete Mix using Tree Based Modeling», IJCA, vol. 6, n.o 5, pp. 18-24, sep. 2010, doi: 10.5120/1076-1406.
D.-K. Bui, T. Nguyen, J.-S. Chou, H. Nguyen-Xuan, y T. D. Ngo, «A modified firefly algorithm-artificial neural network expert system for predicting compressive and tensile strength of high-performance concrete», Construction and Building Materials, vol. 180, pp. 320-333, ago. 2018, doi: 10.1016/j.conbuildmat.2018.05.201.
L. Acuña Pinaud, P. C. Espinoza Haro, I. Moromi Nakata, A. V. Torre Carrillo, y F. García Fernández, «Concreto de alto rendimiento, predicción de su resistencia a la compresión mediante redes neuronales artificiales», TECNIA, vol. 27, n.o 1, Art. n.o 1, jun. 2017, doi: 10.21754/tecnia.v27i1.125.
M. Noman, A. Khattak, Z. Alam, M. Yaqub, y E. Noroozinejad Farsangi, «Predicting the Residual Compressive Strength of Concrete Exposed to Elevated Temperatures Using Interpretable Machine Learning», Practice Periodical on Structural Design and Construction, vol. 29, n.o 4, p. 04024055, nov. 2024, doi: 10.1061/PPSCFX.SCENG-1536.
X. Ying, «An Overview of Overfitting and its Solutions», J. Phys.: Conf. Ser., vol. 1168, p. 022022, feb. 2019, doi: 10.1088/1742-6596/1168/2/022022.
J. H. Friedman, «Greedy function approximation: A gradient boosting machine.», Ann. Statist., vol. 29, n.o 5, oct. 2001, doi: 10.1214/aos/1013203451.
A. Natekin y A. Knoll, «Gradient boosting machines, a tutorial», Front. Neurorobot., vol. 7, 2013, doi: 10.3389/fnbot.2013.00021.
G. Ke et al., «LightGBM: A Highly Efficient Gradient Boosting Decision Tree», en Advances in Neural Information Processing Systems, Curran Associates, Inc., 2017. Accedido: 6 de agosto de 2024. [En línea]. Disponible en: https://papers.nips.cc/paper_files/paper/2017/hash/6449f44a102fde848669bdd9eb6b76fa-Abstract.html
T. Chen y C. Guestrin, «XGBoost: A Scalable Tree Boosting System», en Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco California USA: ACM, ago. 2016, pp. 785-794. doi: 10.1145/2939672.2939785.
M. I. Khan et al., «Effective use of recycled waste PET in cementitious grouts for developing sustainable semi-flexible pavement surfacing using artificial neural network (ANN)», Journal of Cleaner Production, vol. 340, p. 130840, mar. 2022, doi: 10.1016/j.jclepro.2022.130840.
M. Iqbal, Q. Zhao, D. Zhang, F. E. Jalal, y A. Jamal, «Evaluation of tensile strength degradation of GFRP rebars in harsh alkaline conditions using non-linear genetic-based models», Mater Struct, vol. 54, n.o 5, p. 190, sep. 2021, doi: 10.1617/s11527-021-01783-x.
M. Iqbal, D. Zhang, F. E. Jalal, y M. Faisal Javed, «Computational AI prediction models for residual tensile strength of GFRP bars aged in the alkaline concrete environment», Ocean Engineering, vol. 232, p. 109134, jul. 2021, doi: 10.1016/j.oceaneng.2021.109134.
Publicado
Cómo citar
Número
Sección
Licencia
Esta obra está bajo una licencia internacional Creative Commons Atribución 4.0.
El autor permite el uso y difusión de los artículos a los lectores y a la revista, por medio de la licencia Creative Commons CC-BY-4.0 (Reconocimiento 4.0). Esta licencia permite que cualquiera distribuya, remezcle, adapte y rediseñe a partir de un trabajo, incluso con fines lucrativos, siempre que se mencione la autoría de la creación original.
Ruta CC-BY-4.0: https://creativecommons.org/licenses/by/4.0/
