<p>This study proposes the application of machine learning models namely recurrent neural network (RNN), ensemble neural network (ENN), and long short-term memory (LSTM) to predict the factor of safety against sliding, overturning, and bearing failure. A comprehensive dataset of 400 instance which were generated using mean and standard deviation of the input parameters from the previous research and was utilized to train and test these models, with performance evaluated using performance parameters like coefficient of determination (<i>R</i>²), Willmott’s index, Legate and McCabe’s index, root mean square error, scatter index, and mean absolute error. Further analysis, such as rank analysis, regression curve, system reliability evaluation, comparative measure analysis, objective functional criteria, uncertainty analysis, statistical testing, Williams plot, Taylor diagram, error plot, residual curve, and external validation was conducted to determine the most effective model. The results demonstrate that the LSTM model outperforms RNN and ENN in predicting the external stability of a cantilever retaining wall with a <i>R</i>² = 0.998 in training (TR) and <i>R</i>² = 0.997 in testing (TS) phase against sliding failure, <i>R</i>² = 0.999 in TR and <i>R</i>² = 0.998 in the TS phase against overturning and <i>R</i>² = 0.997 in TR and <i>R</i>² = 0.996 in TS against bearing failure. Sensitivity analysis was also performed to check the influence of each input parameters on the output.</p>