Hybrid Semantic Service Matchmaking Method Based on a Random Forest

Abstract

Semantic Service Matchmaking (SSM) can be leveraged for mining the most suitable service to accommodate a diversity of user demands. However, existing research on SSM mostly involves logical or non-logical matching, leading to unavoidable false-positive and false-negative problems. Combining different types of SSM methods is an effective way to improve this situation, but the adaptive combination of different service matching methods is still a difficult issue. To conquer this difficulty, a hybrid SSM method, which is based on a random forest and combines the advantages of existing SSM methods, is proposed in this paper. The result of each SSM method is treated as a multi-dimensional feature vector input for the random forest, converting the service matching into a two classification problem. Therefore, our method avoids the flaws found in manual threshold setting. Experimental results show that the proposed method achieves an outstanding performance.