In order to improve the mining efficiency of target data in a large number of mixed data, a multi-constraint mining algorithm for mixed data based on parallel computing is proposed. The wavelet threshold method is used to denoise the mixed data. A distance parallel algorithm based on the matrix multiplication function in Cublas library is proposed to obtain the distance between the mixed data. Positive and negative association constraints are extended to the denoised mixed data, and the mixed data are clustered based on the constraints and the distance between the data, and the similar data mining is completed according to the clustering results. The experimental results show that the method has good data processing effect and high data mining performance.