With the development of single-cell RNA sequencing technology, it is very important and valuable to supplement and improve the mining algorithm of single-cell RNA data to understand the heterogeneity of single-cell RNA and the precise mechanism of the prevention and treatment of diseases. Machine learning and data mining are the preferred technologies for processing large amounts of data. The multiple fitting regression and Bayes (MFRB) method is a new method that combines multiple fitting regression (MFR) methods and Bayesian decision-making in machine learning. The probabilistic support vector machine (PSVM) method is excellent for data classification and has been widely used and verified. In this study, these two classification methods were used to detect large-scale single-cell RNA data and small-sample unbalanced single-cell RNA data, respectively. The performances of the two algorithms were determined and their classification effects were discussed. A random walking preprocessing algorithm is also used to improve the distribution characteristics of low-quality data. The results show that the two algorithms have good results only for large-scale single-cell RNA data; for small-sample unbalanced data sets, neither of the algorithms effectively classified single-cell RNA data.