摘要
本文叙述了置信度网络与偏最小二乘法（PLS）联用方法的建立，利用置信度网络处理信息的不完整性，利用偏最小二乘法建立预报模型，并预报结果．将此方法运用于实际，根据原子光谱法测定人发中微量元素的浓度值，识别癌症病人，获得较好效果，当元素浓度值78％时，就能获得100％元素浓度值的预报结果。
The application of mathematically sound probabilistic reasoning and pattern recognition techniques in cancer diagnostic based on the concentration of elements in human hair is presented in this paper. Belief networks, which have raised enormous interest in the AI research community, are used to deal with the uncertainty in measurement of the elements concentration. Partial least square(PLS) method is used as a dassifier of cancer and normal samles. The results from the validation shown that the proposed inference scheme is effective in predicting unobserved symptoms. In the PLS analysis, the best classified result can be obtained when ouly 78% symptoms are observed based on the networks inference while the same result may be gained with 100% symptoms observed without the inference.
出处
《计算机与应用化学》
CAS
CSCD
1997年第1期60-64,74,共6页
Computers and Applied Chemistry
基金
厦门市科委科技基金
