摘要
分析了文献中高炉铁水硅含量预测数据,结果表明:评价指标选择决定预测效果的评价结论和模型选择;命中率判据不是评价预测效果的最佳选择;相关系数及其检验是评价预测结果优劣的客观判据;相关系数及其检验、人工预测、随机游走模型可以作为以科学性、实用性和简便性为原则的预测评价判据基准。
Data in previous publications of forecasting silicon content in hot metal of blast furnace are analyzed. The result shows that evaluation of forecasting results depends on the used criteria. Hit ratio is not the most scientific criteria for forecasting evaluation in comparison with correlation coefficient between the measured and forecasted silicon content. The correlation coefficient and its test, judgment of blast furnace operator and random walk model should be considered as benchmark of the evaluation based on rational, practical and simple principle. Normalizing the scientific criteria of forecasting silicon content in hot metal is suggested.
出处
《钢铁研究学报》
CAS
CSCD
北大核心
2007年第5期5-9,19,共6页
Journal of Iron and Steel Research
基金
国家自然科学基金资助项目(59974035)
教育部留学回国人员基金项目(2004-527)
关键词
高炉
硅含量
预测评价
模型选择
blast furnace
silicon content
forecasting evaluation
model selection