Soil Liquefaction Susceptibility Mapping Using a Machine Learning Algorithm: A Case of Guagua, Pampanga

This case study investigates the application of a machine learning algorithm in predicting the soilliquefaction susceptibility of Guagua, Pampanga. The random forest was employed in training thepredictive model using 413 datasets. The variables in the dataset include natural moisture content, D50,plasticity index, SPT N-Value, groundwater level, and fines content. Ethical considerations ensured theprivacy of the data collected from various soil testing centers and DPWH districts in Pampanga. Thetrained model achieved 93% accuracy in predicting soil liquefaction susceptibility after feeding it with thedatasets. In addition, the feature importance revealed that variables such as D50 and groundwater levelhave a more significant influence on the likelihood of soil liquefaction compared to fines content, SPT NValue, PI, and natural moisture content. Based on the model, the locale has 16 barangays that aresusceptible to soil liquefaction, 3 barangays that are not susceptible to soil liquefaction, and 12 barangayswith unidentified soil liquefaction susceptibility. The researchers had concluded that Guagua, Pampangahas a high probability of experiencing soil liquefaction based on the hazard map generated from a webapplication, equipped with the predictive model trained using a random forest algorithm, developed by theresearchers.