In tri-reference point theory, minimum requirement (MR), status quo (SQ), and goal (G) are behavior rules for decision-making subjects, which divide the value into four regions: failure, loss, gain and success. Taking into account the investors' three reference points of investment return, a portfolio selection model is constructed by setting the maximum investors' perceived value in tri-reference point theory as the objective under the premise of meeting the safety-first principle. A particle swarm algorithm that is suitable for solving this optimization problem is designed. On this basis, two types of common structured products in China's financial market are included in two portfolios. Then the purchase behavior of structured products and the superiority of structured products compared to risk-free assets and underlying assets are studied by changing the parameters in tri-reference point theory. The results show that structured products are most favored by investors when both MR and G are relatively high. This is because under these circumstances, structured products can meet the demand of investors to pursue relatively high return on the premise of ensuring safety to make most investors prefer them.