Building an ideal perovskite photovoltaic device requires in-situ losslessly monitoring the transport and recombination characteristics,such as carrier mobility,defect density,depth and capture cross section.In view of this,we establish a new methodology of extracting these characteristics with ultra-high resolutions from a group of opto-electro-modulated electrical transient spectrums.We clearly map the micro carrier dynamic processes in electrical transients by numerical simulation and construct the extra nonlinear relationships between electrical transient spectrums and device characteristics by machine learning.It is found that a certain numbers of electrical transients curves under different steady physical field possess the function of fingerprint for diagnosing the semiconductor characteristics of a solar cell,finally achieving a more than 99% prediction accuracy rate within 5% error tolerance.Using this method,we in-situ losslessly probed the subtle changes of transport and recombination characteristics in Formamidine (FA)-based perovskite solar cells caused by phase transition.The methodology shown in this work owns great inclusiveness and extensibility,which can also be used in other cells or semiconductor devices.