The Minimum Mean-Square Error (MMSE) detector can achieve excellent Bit Error Rate (BER) performance in massive Multiple-Input Multiple-Output (MIMO) systems. However, it involves high complexity large-scale matrix inversion operation with high degree, resulting in very high hardware requirements. To solve this problem, the article proposes a Generalized Weighted-Preconditioned Gauss-Seide (GS) (GW-PGS) iterative algorithm. In this algorithm, an initialization scheme based on preprocessing is first proposed, which speeds up the convergence speed without adding additional complexity. In addition, this paper proposes an adaptive weighting factor scheme. The experimental results show that compared with the traditional GS algorithm, the algorithm proposed in this paper can effectively reduce the BER and computational complexity.