Aiming at the problem of difficulty in selecting access points for hybrid LiFi-WiFi heterogeneous networks in complex indoor environments, a random forest model-based access point selection algorithm for hybrid LiFi-WiFi networks is proposed.The proposed network access point selection algorithm utilizes the channel characteristics of multiple networks, and constructs a training set by simulating different complex indoor environments and collecting the equivalent values of received signal strength and signal-to-noise ratio of users at different locations in different situations, so that the model can adapt to various complex indoor environments.Simulation results show that compared with the traditional network selection algorithm, the average user reachable throughput of this algorithm is improved by about 82%, especially when the indoor situation is more complex, it can be improved by 160%.The algorithm proposed can significantly reduce the number of switching times by 20% compared with other algorithms when the users are moving and the number of accessed users increases.