In this paper, we propose a novel fact-checking (truth-finding) strategy based on machine learning data clustering with the k-means method combined with the silhouette index to determine the optimal value of k, in order to detect the optimal partition of the set of attributes. Such an optimal partition maximizes the accuracy of the truth-finding process without having to explore all possible partitions. Results from our intensive experiments on synthetic and real data show that our approach outperforms those in (Lamine Ba et al., 2015), with a more reasonable computational time cost. Finally, we sketch a way to parallelize a given truth-finding process using the MapReduce paradigm with a view to avoiding the explosion of execution time when the size of the input data increases.
