Cement is an important basic building material that has a significant impact on social production. The rapid detection of cement raw material composition is of great significance for the development of the construction industry. The content detection of Al₂O₃ and Fe₂O₃ in cement raw meal based on near-infrared spectral analysis method is performed. Firstly, the sample set is divided by the combined X-Y distance division method. And the training set is processed by different spectral pretreatment methods. Finally, PLS and SVM are utilized to establish prediction models for NIR data respectively. The predicted results are analyzed and compared and the results show that the NIR analysis method using S-G smoothing pretreatment and PLS modeling has a better detection results. The decision coefficient R² of the Al₂O₃ detection model is 0.895, and the RMSEP is 0.072; the decision coefficient R² of the Fe₂O₃ detection model is 0.732, and the RMSEP is 0.023. The research results provide an effective analytical method for detecting the composition of cement raw materials, promoting the further development of the cement industry.