Abstract: In order to solve the problem of the high mis- judgment rate under the unstable conditions in the transient voltage stability assessment and improve the application of the multivariate decision tree (MDT), a hierarchical cost sensitive- multivariate decision tree (HCS-MDT) assessment is proposed. Based on the spatiotemporal joint expansion of the measurable electrical quantities to construct the features, the cost-sensitive support vector machine (CS-SVM) with the improved empirical risks is used as the internal node classifier of the MDT, and the analytic combined feature stability judgment rule is generated as a visual stability judgment basis, which may effectively reduce the instability misjudgment. The hierarchical self- adaptation (HSA) criterion is integrated into the CS-MDT for the assessment of the transient voltage stability, which effectively guarantees the assessment accuracy while improving the early assessment. The simulation example of the transient voltage stability verifies the effectiveness of the proposed method.