This article presents a unified framework for the early prediction of short-term voltage stability (STVS) using synchronized measurements of voltage, phase angle, and frequency from seven buses in a modified IEEE 14-bus system. A dataset comprising 5000 dynamic simulations was generated in DIgSILENT PowerFactory 2024 SP1 under probabilistic operating conditions and three-phase faults. Random Forest and Conv1D-LSTM models achieved F1 scores of 0.9782 and 0.9783, respectively, for the instability class. An observation window of 300 ms at 120 Hz yielded effective predictive performance while allowing sufficient time for supervisory assessment. Furthermore, an illustrative undervoltage load shedding scheme demonstrated that early warning information can support wide-area monitoring and subsequent control actions during severe contingency conditions.