•Monte Carlo simulation (MCS) solves the sparsity problem for small samples of VHCF.•The proposed dynamic recursive network validates the effectiveness of MCS for enhancing small sample dataset.•Achieving accurate prediction of VHCF for small sample cases and reducing the cost and time. Few machine learning (ML) methods have been applied to the prediction of the highly dispersed and low Very High Cycle Fatigue (VHCF) specimens. The present work attempts to use Monte Carlo Simulation (MCS) to expand the low VHCF dataset and thereby improve the predictive performance of ML models. Moreover, a dynamic recurrent ML model is proposed to be used with the extended dataset. VHCF experiments conducted under three different stress ratios (R = −1, 0.05 and 0.5) and two temperatures (26 °C and 500 °C) show that the use of MCS has significantly improved the predictive capabilities of the ML model, with predictions almost within the scattered band of 4.0 and closer to 2.0, thus solving the challenge of predicting the life of sparse VHCF specimens made of Ti60 alloy, the cost and time of VHCF prediction will be greatly reduced.