Support vector machine-based cascade classification model training method and device suitable for mass spectrum data

The invention relates to a training method and device of a cascade classification model based on a support vector machine and suitable for mass spectrum data. The training method comprises the steps that effective use intervals of original mass spectrum data collected in different states are mapped to sampling intervals, so that features suitable for a support vector machine are extracted to serve as a secondary training set, and the sampling intervals are larger than intervals determined based on the corresponding relation between the actual mass-to-charge ratio and the collection frequency; the secondary training set comprises a first training set and a second training set; using the first training set to perform first training on each category of samples in the plurality of classification categories, and selecting a plurality of classifiers for each category of samples according to the classification accuracy of the trained classifiers; combining the plurality of classifiers by using the second training se