This paper proposes a method of inverse analysis for identification of an impact load based on the data selection approach. Taking the dynamic behavior of structure under the impact load is comparatively easy if many sensors and high precision measuring devices are available. On the other hand, if the dynamic behavior can be identified by a few strain data with lower accuracy, the method has advantages to measuring cost and efficiency. In this study, the application of the proposed inverse analysis method is shown in the designing of the new type of the pitching plaything. In the proposed method, the fuzzy clustering is used for selecting and picking up the strain data containing errors for the neural network learning. As a result, it is shown that the proposed method has an advantage on the inverse analysis for some margin of errors.