In order to realize the accurate prediction of molecular thermal properties in the process of computer aimed molecular design, the molecular boiling point temperature was selected as an important prediction parameter to derive other thermal properties. The existing 6 types of working fluids used in thermal cycles such as alkanes, alkenes, halogenated hydrocarbons, alcohols, ethers, and amines were divided into 16 basic functional groups, and by introducing the topological index EATII to distinguish the isomers. The number of 16 groups occurrences and EATII were taken as input parameters, the boiling point temperature were regarded as the output parameter and will be predicted by constructing an artificial neural network. In the prediction process, the sparrow search algorithm is used to optimize the artificial neural network, and the proposed model sparrow search algorithm- artificial neural network is compared with the existing genetic algorithm-artificial neural network model. The prediction of the present model was contrasted with the experimental data, the results show the average absolute deviations for training, validation and test sets are 0.66%, 0.76%, 0.64%, respectively. The error is significantly reduced, it indicates that the network proposed in this paper can accurately predict the boiling point temperature of the working fluid.