RVM (Relevance Vector Machine) based ATC (Available Transfer Capability) estimator for its preliminary screening is proposed, which estimates the approximate value of ATC without long computing time. Starting from this obtained estimation, we can expect drastic saving of the time required to get accurate ATC. Comparisons are made between the proposed method and other existing methods. The performance of the proposed method is clarified by comparing the results with those of existing learning machines such as MLP (Multi-Layer Perceptron) and SVM (Support Vector Machine). In SVM the same kernel is used as in the proposed system. Not only the parameters in the proposed system but also all the parameters in MLP and SVM are carefully tuned to give the best results.