In order to quickly predict the change in residual oil content after the pyrolysis of oily sludge and to guide the optimization of pyrolysis process parameters, several factors were used as input variables. These included the final pyrolysis temperature, pyrolysis time, heating rate, nitrogen flow rate, oil content, water content, and residue content of the oily sludge. The oil content of the pyrolysis residue was used as the output variable. Predictive models for the oil content in the pyrolysis residue of oily sludge were established using gradient boosting decision tree (GBDT), extreme gradient boosting (XGB), support vector machine (SVM), and random forest (RF) algorithms. Through training and testing on 228 sets of data, the results showed that the determination coellicient R^2 of the test set for the four oil content prediction models (GBDT, XGB, SVM and RF) were 0.871 6,0.866 7,0.835 6 and 0.917 1, respectively. After hyperparameter optimization using the Bayesian optimization algorithm (BOA), the determination coefficient R^2of the test set for the four oil content prediction models reached 0.901 2,0.900 1,0.896 5, and 0.920 4, respectively. The Bayesian optimized random forest (BOA-RF) prediction model had better predictive performanee and could predict the dynamic changes in oil content of the oily sludge pyrolysis residue more accurately.