Abstract: Roasted chicken is an important commercial products widely distributed in China, and is consumed by many consumers. The products, prepared with different spices and other ingredients and different cooking methods, can be contaminated with C.perfringens. In order to better inhibit the growth/survival of C.perfringens in food and improve the scientific application of essential oils, in this paper, we studied the effects of various essential oils from spices on the growth of C.perfringens in cooked chicken breast were investigated. The Oxford cup method and two-fold dilution method were used to determine the in vitro inhibitory activities and minimum inhibitory concentrations (MIC) of cinnamon, black pepper, ginger, fennel and wormwood essential oils on C.perfringens. The essential oils of cinnamon, artemisia and fennel with good bacteriostatic activity were screened out and their inhibitory concentrations against (ATCC13124,C1) were obtained for construction of growth/residual kinetic model of C.perfringens in cooked chicken breast by BP neural network. In order to improve the accuracy of network training, data of including essential oil concentration, the number of strains and the types of essential oils were used as input parameters prior for training; the corresponding surviving counts of C.perfringens (C1 and ATCC13124) in cooked chicken breast were used as the target parameter. After the trained samples were normalized, he trainlm function was used in the network learning parameter, the maximum training times were 5 000, the training step was 0.1, the expected error was 1×10-7 and a feedforward and backpropagation BP-ANN model was constructed. By comparing the correlation coefficient (R2) and the root mean square error of prediction (RMSE) values from the models .The prediction model accuracy and the accuracy index (Af) and bias factor (Bf) of the inactivated growth curve of C.perfringens in cooked chicken breast meat with different essential oil concentrations were determined. The results showed that: As the concentration of essential oil increased, the growth/residual amount of C.perfringens gradually decreased. Among them, cinnamon oil had the lowest inhibitory concentration on ATCC13124 and C1, and the effect was the strongest. The calculated values of the Bf for the BP-ANN model were close to 1 indicating no systematic bias. The prediction accuracy of the BP-ANN model was greater than 0.917 and the degree of fitting was high. For strain ATCC13124, the prediction model worked best for wormwood essential oil, with R2, RMSE, Af and Bf of 0.992, 0.197 CFU/g , 1.015 and 1.000, respectively. In the case of the isolated strain C1, the prediction model performed best for cinnamon essential oil, with R2 of 0.976, RMSE value of 0.271 CFU/g, Af and Bf of 1.022 and 1.000, respectively. The robustness of the model was verified by the verification set. The results showed that R2 is above 0.917, and the RMSE was between 0.200-0.640 CFU/g. The predicted value and the measured value were better. In summary, the BP-ANN model can be used to rapidly obtain the optimal inhibitory concentration for different essential oils and quantitatively predict the survival of C.perfringens. This work can be used to assess the food safety hazards of C.perfringens in cooked chicken breast and provides a theoretical basis for the control of C.perfringens during meat processing.