By increasing the degree and sophistication of automation, e-marketplaces will become much more efficient and transparent, and hence more widely adopted by organizations. Negotiation is one of main activities conducted in e-marketplaces, and adaptive negotiation agents can be applied to improve the effectiveness of B2B e-marketplaces. Classical negotiation models have limited use in modern e-marketplaces because these models often assume that complete information about the negotiation spaces is available. This paper illustrates the design and development of adaptive negotiation agents for e-marketplaces. These agents are empowered by the Bayesian learning mechanisms so that they can gradually acquire negotiation knowledge based on their previous encounters with the opponents. Our preliminary experiment shows that the proposed probabilistic negotiation decision making mechanism and the associated data mining approach is effective and efficient in simulated e-marketplaces.