The proposed model labels the negotiation history data automatically by making full use of the implicit information in negotiation history. Then, the labeled data become the training samples of least-squares support vector machine that outputs the estimation of opponent’s utility function. After that, the self’s utility function and the estimation of opponent’s utility function constitute a constraint optimization problem that will be further figured out by genetic algorithm. The optimal solution is the counter-offer of oneself. Experimental results show that the proposed model is effective and efficient in environments where information is private and the prior knowledge is not available.