Abstract:  Coal blending plays a key role in the production of a coking plant. In the iron and steel industry，CSR and CRI indexes of coke at high temperature are of particular importance. In order to improve traditional coal blending process that relies on expert experience，the idea of predicting coke quality based on raw coal indexes has been put forward. This article aims to analyze some coke quality prediction methods at home and abroad and improve the existing deficiencies. Taking Phase I and II projects of Xilaifeng coking plant as an example，this article conducts intelligent analysis, using SVR model，on CSR and CRI indexes of coke. By showing the relationship between training sample proportion and prediction accuracy，it further points out research orientation of coke quality prediction work.

Key words:  , Support vector regression（SVR）, Coke strength after reaction（CSR）, Coke reactivity index（CRI）