S. González, J. Vega, A. Murari, A. Pereira, S. Dormido-Canto, J.M. Ramírez and JET EFDA contributors Completely automatic techniques to estimate and validate L/H transition times can be essential in L/H transition analyses. The generation of databases with hundreds of transition times and without human intervention is an important step to accomplish (a) L/H transition physics analysis, (b) validation of L/H theoretical models and (c) creation of L/H scaling laws. An entirely unattended methodology is presented in this article to build large databases of transition times in JET using time series. The proposed technique has been applied to a dataset of 551 JET discharges between campaigns C21 and C26. The prediction with discharges that show a clear signature in time series is carried out through the locating properties of the wavelet transform. It is an accurate prediction and the uncertainty interval is ±3.2ms. The discharges with a non clear pattern in the time series use an L/H mode classifier based on discharges with clear signature. In this case, the estimation error shows a distribution with mean and standard deviation 27.9ms and 37.62ms respectively. As an additional original development, the level of significance of the predictions can be used to determine a variable time segment around the transition, in which there is a guaranteed probability to find the time of the L to H transition. Its temporal length correlates with the input power: increasing the input power, the plasma moves to a clearer H-mode regime, reducing the length of this guaranteed probability confidence interval.