Wireless channel propagation loss prediction method driven by measured data

The invention discloses an actually measured data-driven wireless channel propagation loss prediction method, which comprises the following steps of: for a scene to be measured which has an actually measured condition and is sufficient in actually measured data volume, establishing an artificial neural network as an initial prediction network, and training the initial prediction network by adopting actually measured data; the method comprises the following steps: for a to-be-measured scene which has actual measurement conditions but is insufficient in actual measurement data volume, adopting an artificial neural network of double weight neurons as an initial prediction network, carrying out pre-training by using network training simulation data, and then carrying out optimization training by using network training actual measurement data; and for a to-be-tested scene without actual measurement conditions, executing an inter-scene network parameter migration module, adopting an artificial neural network of dou