Abstract: Material data is featured as small sample, high noise, high dimension, complex associations and rich expert knowledge. A dataset with 410 data points was collected. Composition condition and property data is included in the dataset. The copper alloy state symbols were recoded refer to the one-hot coding method. First, network structures (input - layer1- layer2 - output) of strength and conductivity models were optimized and set as 21-55-70-1, 21-50-65-1 respectively. Based on the optimized network structure, expert knowledge was quantitatively characterized and embedded into neural network loss function. For example, strength will increase with the increase of harden level, and this rule can be quantitatively expressed with Spearman score. The Spearman scores of 6 rules were added to the loss function. Then, an expert-augmented machine learning model was trained and network weights were optimized with genetic algorithm. Every time the network weights were updated, orthogonal data was generated to evaluate the consistency between the output data and expert knowledge. The Spearman correlation coefficients of model input-output data and expert knowledge are above 0.98. The R2 scores of both strength and conductivity models achieved on the test set are above 0.90. A multi-objective optimization with genetic algorithm was carried out based on the composition, condition, strength and conductivity models. After 100 generations of iteration, Pareto optimal solutions were found and experimentally validated. The results show that the tensile strength can be as high as 637MPa while the conductivity is 77.5%, and the conductivity can be as high as 80.2% while the tensile strength is 600MPa. Relative errors between experimental values and predicted values were less than 5%. The microstructure images show that coarse second phases are found in the as-cast structure. After solid solution, cold deformation and aging process, the grains were stretched and the coarse second phases were redissolved and redistributed. Those precipitated particles distributed along grain boundary tend to have a lower strength and conductivity. According to the experiment result, Mg, Ti elements are detrimental to the increase of strength, while Fe, Sn elements can effectively increase the strength. The analysis also shows that Fe element has less influence on conductivity and Sn element has a great influence on conductivity.