Various methods for detecting malaria have been developed in recent years, each with its own set of advantages. These methods include microscopic, antigen-based, and molecular-based analysis of blood samples. This study aimed to develop a new, alternative procedure for clinical use by using a large data set of surface-enhanced Raman spectra to distinguish normal and infected red blood cells. PCA-LDA algorithms were used to produce models for separating P. falciparum (3D7)-infected red blood cells and normal red blood cells based on their Raman spectra. Both average normalized spectra and spectral imaging were considered. However, these initial spectra could hardly differentiate normal cells from the infected cells. Then, discrimination analysis was applied to assist in the classification and visualization of the different spectral data sets. The results showed a clear separation in the PCA-LDA coordinate. A blind test was also carried out to evaluate the efficiency of the PCA-LDA separation model and achieved a prediction accuracy of up to 80%. Considering that the PCA-LDA separation accuracy will improve when a larger set of training data is incorporated into the existing database, the proposed method could be highly effective for the identification of malaria-infected red blood cells.