Abstract: Predicting the failure time of landslides remains a significant challenge in landslide research. The inverse velocity method is widely employed for landslide failure time prediction due to its effectiveness and practicality. The accurate identification of the onset of acceleration(OOA)critically influences the predictive results of this method. To address this,we present a straightforward approach for identifying OOA based on the exponential moving average technique,commonly used in economics. The method involves three steps: (1)analyzing absolute velocity values; (2)defining a trend change index ω and identifying acceleration trend zones using a moving time window technique; and(3)performing linear regression analysis on inverse velocity within acceleration zones,with OOA points determined based on correlation coefficients. Applied to the Qubuga landslide case,the model demonstrated effective identification of OOA points,with correlation coefficients exceeding 0.8 and prediction errors of less than four days,confirming the method's practical utility.