<p>Real-time prediction of traffic conflict risk within a connected vehicle environment is a crucial aspect of intelligent transportation system. This proactive real-time traffic conflict risk prediction allows stakeholders to predict dangerous traffic interaction and prepares evasive maneuvers in advance. However, current research mainly focuses on predicting the traffic conflict risk based on vehicle kinematic characteristics, without considering built environment characteristics and land use characteristics. Addressing the notable lack of prior research, our study introduces real-time prediction of longitudinal traffic conflict risk based on the four real-time models for this purpose: Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) Network model. These models are evaluated using connected vehicle data, which includes time series vehicle kinematic characteristics, built environment characteristics and land use characteristics. Diverse features regarding these three characteristics are extracted from the dataset and incorporated into these four models. Results indicate that the LSTM model surpasses the other models in terms of several prediction performance indicators, underscoring the effectiveness of LSTM in predicting longitudinal traffic conflict risk. Results also show that the prediction performance is enhanced by integrating built environment characteristics and land use characteristics. Moreover, recursive feature elimination analysis shows that speeding, speed, acceleration and hospital are the most important influential factors in predicting longitudinal traffic conflict risk, followed by intersection, land-use diversity and tree area proportion. This research contributes to the development of intelligent mobility systems by demonstrating how connected-vehicle data and deep learning can facilitate data-driven safety monitoring and support smart infrastructure design.</p>