Abstract: A method for predicting cold gas content in a melter-gasifier was proposed based on entropy-weighted fuzzy C-means clustering and partial least squares(PLS).In the modeling process,an entropy-weighted fuzzy C-means clustering algorithm is used to get the clustering result of burden calculation reports according to the consumption of raw materials at first.Then,different prediction models are built based on a PLS algorithm for various cluster types.The real field data of cold gas content from Baosteel COREX-1^# were used for verification.It is shown that the method can build the prediction model of COREX cold gas content effectively,and has an advantage in prediction accuracy.