Based on the interdependency relationship among projects, the paper analyses risk factors in the project portfolio network via the random walk algorithm. Sustainability is one of the most important challenges of the project and portfolio management. This paper analyses the interdependencies among projects in a portfolio from the perspective of sustainable development and builds models to measure the relationship among risk factors via the Multidomain matrix (MDM) method. Using the interdependency relationship among projects and potential relationship between different risk factors as inputs, the paper builds the model of portfolio risk network to predict the risk in the project portfolio via a random walk algorithm. Because the random walk is a personalized recommendation algorithm, so our proposed method can achieve an accurate prediction of portfolio risk through predicting the risk factors and their probabilities in the portfolio. Our method can also help project managers to rank these risk factors in the portfolio through distinguishing the most concerned risks.