Fast and accurate state of health (SOH) prediction of lithium-ion batteries is an effective guarantee for the safety and reliability of their operation process. However, the majority of existing research focuses on enhancing the SOH prediction accuracy adopting advanced algorithms, and seldom considers the improvement of the prediction efficiency. For this reason, this study first analyses and processes the acquired battery aging data. Then, the charging time of the partial charging voltage profile is treated as the health feature, which circumvents the complex and cumbersome feature extraction process. Subsequently, an efficient battery SOH prediction model is developed, which exploits adaptive learning particle swarm optimization to search for the hyperparameters of the least squares support vector machine. The validation results indicate that the proposed method enables fast training and prediction of the model in less than 1 s. Moreover, the maximum absolute error of SOH prediction on all batteries is less than 2%, which validates the feasibility and applicability of the proposed method.