Abstract: 【Purpose/significance】The network society is full of many negative network public opinions. The risk classification and judgment of negative network public opinion is of great significance to improve the ability of network governance and the effectiveness of network social governance.【Method/process】The risk index system of negative network public opinion is constructed, and the en⁃tropy weight method is used to calculate the weight of risk index; the grey weighted information correlation degree is calculated based on the weighted GRA model, and on this basis, the k-means clustering algorithm is used to construct the risk classification scheme of negative network public opinion, and then the risk prediction of negative network public opinion is carried out.【Result/conclusion】The results of empirical analysis show that the model has strong objectivity and high reliability, which can provide effective decisionmaking basis for accurate response of negative network public opinion risk.【Innovation/limitation】From the perspective of Informa⁃tion Association, it provides a new research framework for risk classification and prediction of negative Internet public opinion, but the typical case database needs to be improved.