The invention discloses a millimeter wave MIMO user incremental cooperative beam selection method based on wide learning, which comprises the following steps that: aiming at the downlink beam selection problem of a multipoint cooperative millimeter wave large-scale MIMO scene, each user collects downlink wide beam response and transmission narrow beam response, trains a local wide learning network, and transmits the downlink wide beam response to the local wide learning network; and beam selection is carried out based on the predicted narrow beam response. Furthermore, the training problem of the local network of each user is modeled into a distributed optimization problem with consistency constraint, and effective sharing of training data can be realized by utilizing D2D communication between adjacent users. Furthermore, an incremental updating mode of the local network of the user in the cooperation mode is designed, so that the training complexity of the network can be effectively reduced. According to the method, the capability of mining the relationship between the multi-base-station wide beam response and the transmission narrow beam response under the small sample condition of distributed wide learning is fully utilized, and low-complexity and low-overhead beam selection of the fast time-varying scene multipoint cooperation millimeter wave large-scale system can be realized.