The invention discloses an expressway road condition identification and prediction method based on floating vehicle data, and the method comprises the following steps: carrying out the data preprocessing based on the original GPS track data, based on a speed calculation method for grid division results, calculating a travel vehicle speed by using an Euclidean distance between track points and travel time as an instantaneous speed of a vehicle at an updating point; based on the traffic flow parameters in the speed calculation grid, calculating the traffic flow parameter value in each unit by taking the space grid and the time grid as units; dimensionally reducing traffic state parameters based on principal component analysis, simiplifying the dimension of data and abandoning irrelevant features in the data; based on k-means traffic state clustering anylsis, identifying different traffic states; and constructing characteristics of different time scales, and establishing each traffic state quantity prediction model based on the long short-term memory neural network LSTM. The expressway traffic state can be accurately identified, and the evolution trend of the expressway traffic statecan be predicted.