The invention belongs to the field of intelligent traffic systems, particularly relates to a traffic path recommendation method, system and device based on parallel integrated learning and aims to solve a problem that an optimal vehicle driving path cannot be recommended in the prior art. The method comprises steps of defining a feasible region based on the traffic demand information, acquiring historical data, and dividing the historical data into different historical data subsets according to road sections; obtaining the prediction result by adopting a road section prediction sub-model respectively, obtaining an optimal hyper-parameter set of the model, updating the artificial simulation scene, drawing a traffic jam map of a feasible planning area, and obtaining a recommended driving traffic path; training the road section prediction sub model through a parallel integrated learning method. The method is advantaged in that an integrated model is adopted to predict future demands, thefuture demands are taken as the main basis for constructing an artificial simulation scene, the information is updated synchronously and dynamically, big data is generated through small data, small knowledge is extracted from the big data, so traffic efficiency of roads is effectively improved, the congestion condition is alleviated, and the travel consumption time is reduced.