Housing building projects require careful project management because of their lengthy lead times, significant investment requirements, and high-risk nature. Aimed at effective management and risk assessment of engineering project construction, a risk management model for the entire process of project engineering is established. Risk information on engineering construction projects is obtained through case studies and relevant literature data, and key risk factors are screened using big data technology. Considering the complexity and nonlinearity of risk factors in engineering project construction, a feedforward model (BP) is adopted to solve the risk management model and achieve project risk prediction. Meanwhile, considering that traditional BP models are affected by initial parameters during the training process, they are prone to local convergence problems. Innovatively introducing a Sparse Search Algorithm (SSA) to optimize the construction of the SSA-BP engineering risk prediction model, achieving project risk management and evaluation. In the risk level prediction of risk factors, the Particle Swarm Optimization-Back propagation (PSO-BP) has a large error from sample 15 to sample 30, and the average prediction accuracy of the risk factor level is 73.65%, while the average prediction accuracy of SSA-BP model is 92.65%. In the project risk factor prediction, the average prediction accuracy of the SSA-BP model and PSO-BP model are 91.68% and 82.69%, respectively, which shows that the SSA-BP model has better risk management ability. The SSA-BP model exhibits higher precision and accuracy, improving the ability of engineering project risk management. In addition to offering trustworthy tools and procedures for decision-making in linked sectors, research provides a significant technical reference value for risk management in building projects.