This paper presents a Deep Multi-View Spatial-Temporal Network (DMVST-Net) for predicting taxi demand, aiming to improve resource allocation and reduce traffic congestion in smart cities. The proposed model integrates three views: temporal, spatial, and semantic, to capture complex nonlinear spatial-temporal relationships. The temporal view uses LSTM to model sequential dependencies, the spatial view employs a local CNN to capture spatial correlations among nearby regions, and the semantic view constructs a graph to model functional similarities between regions. Extensive experiments on a large-scale taxi demand dataset from Didi Chuxing in Guangzhou, China, demonstrate the effectiveness of the proposed method, outperforming state-of-the-art methods in terms of Mean Average Percentage Error (MAPE) and Rooted Mean Square Error (RMSE). The model's robustness is further validated through performance analysis on different days of the week and sensitivity to sequence length and input size.This paper presents a Deep Multi-View Spatial-Temporal Network (DMVST-Net) for predicting taxi demand, aiming to improve resource allocation and reduce traffic congestion in smart cities. The proposed model integrates three views: temporal, spatial, and semantic, to capture complex nonlinear spatial-temporal relationships. The temporal view uses LSTM to model sequential dependencies, the spatial view employs a local CNN to capture spatial correlations among nearby regions, and the semantic view constructs a graph to model functional similarities between regions. Extensive experiments on a large-scale taxi demand dataset from Didi Chuxing in Guangzhou, China, demonstrate the effectiveness of the proposed method, outperforming state-of-the-art methods in terms of Mean Average Percentage Error (MAPE) and Rooted Mean Square Error (RMSE). The model's robustness is further validated through performance analysis on different days of the week and sensitivity to sequence length and input size.