This paper presents an ensemble machine-learning approach to monitor the blood volume decomposition state for early hypovolemia detection. Hypovolemia is one of the major causes of preventable deaths in trauma cases. The proposed algorithm discriminates hypovolemia from normovolemia and further classifies hypovolemia into relative and absolute hypovolemia. The algorithms for blood volume classification are analyzed by extracting 13 distinct features from multi-modal physiological signals including Photoplethysmogram, Electrocardiogram, and Seismocardiogram. We compared different Machine Learning classifiers for the multi-class classification problem. We have validated our algorithm on a publicly available dataset collected from six animals undergoing normovolemia, relative hypovolemia, and absolute hypovolemia conditions. The best-performing algorithm is the Artificial Neural Network- (ANN) for the normovolemia/hypovolemia classifier with an accuracy of 93.2% and an F1-score of 0.97. For the absolute/relative hypovolemia (AH-RH) classifier, Long Short-Term Memory offers an accuracy of 95.1% and an F1-score of 0.97. The proposed classifiers outperform the state-of-the-art algorithms and achieve the highest accuracy and F1-score, serving as a potential decision-support tool to observe blood volume decomposition state and help develop context-sensitive hypovolemia therapeutic strategies.