Robot Trust Evaluation and Fault Detection using Weighted Trust Rank and AdaBoost Framework

Web spam degrades the reliability of search engine results by manip- ulating ranking algorithms to raise webpage positions artificially. TrustRank is a popular link-based anti-spam method, its ability to demote highly ranked spam webpages is limited by its exclusive dependence on out-link data. This paper proposes a Weighted Trust Rank (WTR) approach that incorporates link weight into the conventional TrustRank framework in order to overcome this limita- tion. The proposed approach makes use of several link-splitting and accumu- lation strategies, including equal splitting, logarithmic splitting, and Maxshare accumulation, to enhance spam webpage detection. To evaluate the probability of spam content, a Weighted Spam Mass (WS M) measure is also introduced. A robust ensemble model for identifying influential and spam webpages in web graphs is created by applying the AdaBoost method to integrate multiple weak classifiers through iterative reweighting, thereby further improving classifica- tion performance. The suggested WTR approach consistently beats traditional TrustRank, Anti-TrustRank, and Weighted Anti-TrustRank in terms of preci- sion, recall, and accuracy, according to experimental results on two sample web graphs and the WEBSPAM-UK2006 dataset. Additionally, the WTR frame- work may be expanded to multi-robot systems for fault detection and trust as- sessment, showcasing its flexibility in improving dependability and collabora- tive decision-making.