The short-term voltage stability (STVS) of a multi-infeed high-voltage direct current (MIHVDC) system is controlled by complex dynamic reactive power interactions that have multifactor coupling and strong time-varying characteristics. However, because of the complexity in decoupling these interactions, traditional analytical methods cannot accurately quantify their effects on stability. The STVS analysis of MIHVDC systems involves several challenges, such as high-dimensional, time-varying, and nonlinear characteristics. To address these challenges, this paper proposes an STVS evaluation method based on a multi-time-scale (MTS) interaction mechanism and time-series data mining. First, a multi-timescale interaction mechanism is proposed on the basis of dynamic responses. The short-term voltage instability process can be divided into three stages, namely, power flow transfer, dynamic reactive power regulation, and hierarchical cascade electromechanical interactions. This clarifies the accumulation and propagation mechanisms of reactive power deficiency at different timescales. Second, this paper proposes a time-series data mining method based on shapelets to solve problems in traditional STVS analysis. However, the dynamic response caused by commutation failure of the HVDC system interferes with the accuracy of time-series data mining. To solve this problem, this paper presents the optimization of the method by combining it with the dynamic mechanism and proposes the synchronous cluster of shapelets (SynCShapelet) method. In addition, the physical relationship between the shapelet and critical operating point of the dynamic load is elucidated, thus addressing the black-box problem and low confidence of machine learning methods. As demonstrated via a case study, SynCShapelet can predict the instability of the system by detecting features and incorporating the MTS interaction mechanism to preliminarily assess the path of the short-term voltage instability. In the application scenario of the STVS of MIHVDC, the proposed method provides a theoretical basis and technical support for the STVS evaluation and control strategy.