基于加权启发式搜索的鲁棒性信任路径生成

Robust trust path generation based on weighted heuristic search

在社交媒体中,信任传递在用户交互关系的建立上发挥着至关重要的作用.实际应用中,通常将信任传递过程应用于推荐系统来预测起始用户对特定目标用户的信任程度,从而更好地作出下一步决策.选择的信任路径是否较优与预测的准确性息息相关.针对路径长度和信任值在整条路径上的值分布,提出一种新的加权启发式搜索信任预测模型.该模型将改进的经典A^* 算法应用于信任网络进行路径寻找,其中,改进的A^* 算法在寻路过程中使用了二次启发并将启发函数设置为筛选条件进行路径筛选.该模型最终得到的信任路径具有相对较好的鲁棒性,相对提高了预测的准确性,而且在信任累加计算中融入了信任的衰减.最后,通过对比实验验证该模型的有效性,并分析了参数的变化对结果预测的影响.

Trust propagation plays an important role in establishing interactions between users.In practical application,it has been often used in recommendation system to help provide users more valuable and reliable information by collecting advice from other trustworthy users.It can predict the extent of trust from a source user to a specific target user and make further decisions more accurately.The quality of trust path is greatly relevant with the accuracy of predicted results.Thus,based on the length of trust path and the distribution of values on the whole path,we proposed a new trust prediction model called Weighted Heuristic Search Trust(WHST)model in this paper to find high quality paths for prediction.Firstly,the model applies Breadth-first Search(BFS)in path-finding to find all shortest paths which can reach the target user.Then,it employs the improved classical A*algorithm in path-selection to choose the optimal one from them.In this process,it uses two times heuristic function and sets heuristic functions as selection conditions to exclude invalid paths.Only the path with high trust and good robustness could be selected as the optimal path.We combined trust decay with trust aggregation to reflect the fact that trust is weakened in varying degrees as the path length extends.Finally,we evaluated the effectiveness of WHST by carrying out several comparison experiments.The influence of parameter adjustment exerted on the experimental results was discussed as well.