With the application of the thermal manikin and fire manikin, mass data are produced during the experiment research in the clothing engineering area. The advantages of the larger samples can not be revealed using the conventional analysis method. Thus, in the paper, by using the Clementine software, data mining method is used to explore the data produced by flash fire experiment. The decision tree method and the neural net method are used to determine the key influence factors of the thermal shrinkage, which are then used in Kohonen cluster to divide protective clothing into different parts. Research shows that heat flux is the most important factor to the degree of shrinkage and the arm and leg are the key parts to be protected. It suggested that data mining is an effective tool to explore the character and function of the protective clothing.