The invention provides a neural network-based cable life prediction method. The method comprises the steps of training a neural network by adopting temperature, humidity, pressure and voltage information in a cable application environment as variables; after the neural network for cable life prediction is established, inputting temperature, humidity, pressure and voltage information of a to-be-detected cable environment; and performing calculation by the neural network to obtain a life prediction value of a cable, thereby performing assessment on a current usage condition and residual life of the cable and providing a reference for cable replacement. A cable life prediction model is built by adopting the neural network, so that the deficiencies of a conventional cable life detection method are overcome, the cable structure is not destroyed, an assessment result is small in deviation, and a detection result is accurate. The reference is provided for the cable replacement, tedious cable replacement processes are reduced, material waste can be reduced, costs can be lowered, accident hidden danger is reduced, and disasters such as a fire and the like due to the cable are reduced.