The fifth generation of mobile communications system (5G) will be deployed in 2020 and provide diverse communication capabilities, but the promising future of beyond 5G is surely necessary for the communication requirements incurred by fast growing information technology in the next decade. Among several roadmaps toward beyond 5G, wireless big data (WBD) plus artificial intelligence (AI) based communication technology, which covers physical layer, network layer and application layer, is considered as one of the most promising ways. Along with this thought, some emerging research works have been published, which further stimulate more researcher to pay more attention in this area. This paper introduces three interesting works toward this aim, which covers channel modelling, huge access, and network topology design. The channel modelling part starts with the feasible ways to apply machine learning to wireless channel modelling, and presents the prevailing methods in parameter estimation, channel multipath clustering, which is of great importance for future research. The huge access part focuses on fractal phenomenon and its possible applications in wireless networks. After introducing the basic concept, this part investigates the maximum capacity fractal D2D social networks. The network topology design part proposes an interesting topic, whose motivation is to utilize the dynamic mobility features of mobile users to decrease the wireless resource consumption in ultra dense networks (UDN). In summary, these three works are considered as promising topics in beyond 5G, which combines wireless big data analysis and may shed light on related future research.