Abstract: Independent fission yields are essential nuclear data in the field of nuclear reactor engineering. Evaluation of independent fission yields is of great significance. Chain yields describe the mass distribution of fission products and constitute a crucial part of post-fission data. This study aims to collect and analyze the latest experimental data on chain yields, make necessary corrections to the data and evaluate the recommended value of each mass chain. The parameters of the Katakura five-Gaussian model were optimized based on recommended values to calculate for mass chains lacking experimental data. Chain yields of different fissioning systems were generated based on experimentally recommended values and theoretical model calculation results. In the evaluation of chain yield experimental measurement data, more than 21 000 data points from different sources were first collected; then the data were processed for relative measurements and measurement uncertainties were adjusted. The weighted average of chain yield measurement data of a certain mass chain was calculated by the inverse variance weighted method, and the recommended values for different mass chains were evaluated for a certain fissioning system. Based on the evaluated recommended values of chain yield experimental data, the gradient descent method was used to optimize the parameters of the Katakura five-Gaussian model and the chain yields were calculated for mass chains lacking experimental data. The framework of optimization was constructed based on TensorFlow developed by Google. The generation of chain yields for different fissioning systems was completed by combining the recommended values of experimental data and the calculated data of the optimized five-Gaussian model. The calculated values were normalized with the experimental recommended values to ensure data smoothness. The chain yields of different actinide nuclides were generated. Taking the fission of ²³⁵U, ²³⁸U and ²³⁹Pu induced by thermal neutrons, fast neutrons, and 14 MeV neutrons as examples, the results were compared with three sets of chain yield data published separately by the USA and Europe. The reduced χ² between chain yield data and experimental data were calculated for different fissioning systems for different chain yield data sets. The reduced χ² results calculated based on the chain yield data of this work are closest to 1, which is better than the reference data libraries. The chain yield generation method in this study is correct, and the results are reasonable and reliable, which is beneficial to the subsequent independent fission yield evaluation.