To solve the problem of low fault location accuracy and vulnerable to transient component after fault, a data-driven fault location method for transmission lines is proposed. The method first build based on system condition and fault location of power system operation data set. Then based on the maximum correlation - the least redundancy in the centralized data mining the implied relationship between key characteristics and fault location. Extraction and fault location is the largest and related variables between minimum redundancy of key information. Build and fault location at the heart of the most relevant variables set. The last application of curve fitting technology acquisition operation variables and the mathematical relationship between the fault position. Integrated multiple characteristics information accurate fault location results are given.Simulation results show that compared with traditional impedance-based fault location method, the proposed method has higher fault location accuracy and good adaptability to transient components after fault, which is a new accurate and fast fault location method for transmission lines.