Trust Inference Path Search Combining Community Detection And Ant Colony Optimization

Finding trust inference paths for unfamiliar users in online social networks is a fundamental work of trust evaluation. Most existing trust inference path search approaches apply classical brute-force graph search algorithms, which leads to high computation costs. To solve this issue, we propose a trust inference path search approach combining community detection and ant colony optimization. First, the singular value decomposition signs method is utilized to process the trust relationship matrix in order to discovery the trust communities. Then, by taking the communities as different colonies, we use the ant colony optimization to find the optimal trust inference path along which the witness has the maximum deduced referral belief. The released pheromones in previous trust inference path searches help subsequent searches to reuse previous experience and save path search costs. Comparative experiments show that the proposed trust inference path search approach outperforms the existing ones on path search efficiency and trust inference accuracy.