Abstract: The competition among global major powers and non-state actors for regional discourse power and critical mineral resources is profoundly reshaping the investment environment and security landscape in sub-Saharan African countries. Investigating the evolution mechanism of security risks concerning Chinese enterprises and the core driving role of Chinese investment holds significant practical importance. Based on data from 484 security incidents involving Chinese interests in sub-Saharan Africa from 2013 to 2024, this study integrates multidimensional time-lagged variables to construct a random forest model, e.g. investment stock, armed conflict, and unemployment rate. This model captures the nonlinear relationships between security risks related to Chinese interests and their influencing factors, while the SHAP framework is employed to analyze the contribution mechanisms of key variables. The findings reveal that: ① Chinese investment serves as the core endogenous variable driving security risks related to Chinese interests, exhibiting characteristics of “dual pathways, time-lagged effects, and scale thresholds”. Specifically, in the short term, increased investment stock raises risks due to heightened exposure, whereas in the long term, it continuously drives risks through structural cumulative effects, with a clear inflection point in investment scale. This provides empirical evidence for an “investment scale curse” under the resource curse framework in cross-border investment.② Risk generation follows a logic of “time-lagged driven and threshold response”. Within the sample of this study, when the number of conflict events exceeds 200 and the unemployment rate surpasses 15%, the system enters a high-risk state and triggers a risk surge. ③ Security risks related to Chinese interests evolve spatiotemporally through three stages: “local diffusion → concentrated escalation → high-level fluctuation”. High-risk areas remain stable over the long term, with spatial agglomeration continuously intensifying. Theoretically, this study repositions Chinese investment as an endogenous driver of risk. Methodologically, it integrates machine learning with time-lagged mechanism analysis to develop a risk assessment framework with early-warning capabilities. This framework can provide decision-making support for Chinese enterprises in sub-Saharan Africa to implement dynamic risk prevention and control measures.