Adaptive Learning Model for Predicting Negotiation Behaviors through Hybrid K-means Clustering, Linear Vector Quantization and 2-Tuple Fuzzy Linguistic Model

Abstract

Acknowledged System of systems (SoS) manager has no direct control over contributing systems yet they deliver capabilities required to meet the purpose of the SoS operating in an interdependent environment. Forming a joint capability by contribution requires both persuasion and negotiation between the SoS coordinator and each individual system. Negotiation here involves multiple parties, and multiple issues. The challenges here include predicting an opponent system's negotiation behavior and reaching a near optimal negotiation outcome based on the following three issues: performance demands made by the coordinator, monetary benefits in lieu of effort required, and the deadline assigned to prepare for the final SoS participation formation. The negotiation framework proposed involves an unsupervised clustering of the difference between offers from both parties on all three issues by clustering techniques. The clustering results so far indicate the presence of four prominent behaviors: selfish, semi-cooperative, opportunistic, and extremely selfish. The clustered data was used to train both a radial bass function network (RBFN) and a linear vector quantization network (LVQN) to predict future offers. Once trained, the SoS manager uses 2-tuple fuzzy linguistic representation model for decision making. A time-dependent strategy is used to generate a new offer by SoS manager to the systems.