Existing guidance method of demand-side participating in flexible interaction do not fully consider the influence of social infor?mation such as preferential information on load, resulting in a large deviation between the actual role of the guiding cues and the expectedresults. In view of such situation, the Attention-LSTM model is used to forecast the load adjustment under the effect of social information,such as guiding cues. Coupon guidance method could be determined with this model. Take time of use tariff and coupon as an example,coupon guidance method considering the influence of preferential information is proposed, to guide consumers to change their electricityconsumption behaviors for less cost of purchasing electricity. And on this basis, a decision-making model of purchase-sale strategy consid?ering the influence of preferential information in multi-level electricity market is established, to maximize the revenue of electricity retail?ers and verify the feasibility and effectiveness of the proposed guidance method, which is solved by particle swarm optimization algorithmand CPLEX, to obtain the optimal guidance strategy and purchase strategy of electricity retailers. An example based on Australian datashows that the electricity purchasing and selling optimization model of electricity ratailers base on coupons guiding can effectively guidecustomers to actively participate in the interaction.