论文部分内容阅读
针对自动化协商问题,提出一种基于主动学习算法的对手协商偏好学习方法.在该方法中,协商过程表示为建议序列,将建议序列映射到出价轨迹特征空间,建立训练样本集.在激烈竞争的电子商务环境中,样本标记的成本较高,引入主动学习算法后,在预算范围内,提高了对手协商偏好预测的精度.实验数据表明,该方法能在少量有标记训练样本下获得良好的预测能力,减少了协商回合数,提高了协商总效用.
Aiming at the problem of automatization negotiation, this paper proposes an adversarial negotiation preference learning method based on active learning algorithm, in which the negotiation process is expressed as a proposed sequence, the proposed sequence is mapped to the bidding trajectory feature space, and a training sample set is established. In the highly competitive In the e-commerce environment, the cost of sample tags is high, and the introduction of active learning algorithm improves the accuracy of adversary negotiation preference prediction within the budget.Experimental data show that this method can get a good prediction under a small number of labeled training samples Ability to reduce the number of rounds of consultations, improve the overall effectiveness of the negotiations.