等级效标分数的概率神经网络预测方法研究  

Predicting Grade Criterion Scores with Probability Neural Networks

作　　者：余嘉元[1] 

机构地区：[1]南京师范大学心理学系,南京210097

出　　处：《心理科学》2007年第3期666-667,共2页Journal of Psychological Science

基　　金：全国教育科学"十五"规划项目教育部重点课题(DBB010511);国家汉办"十五"科研规划项目(HBK01-05/088)

摘　　要：针对基于统计学的等级效标分数预测存在的问题,提出了运用概率神经网络进行等级效标分数预测的方法。在20种条件下进行了计算机模拟实验,包括测验分数为单变量和多变量,以及各种水平的白噪声干扰条件,结果表明在测验分数为多变量的情况下,或者在有白噪声干扰的条件下,运用概率神经网络方法可以比统计学方法更好地对等级效标分数进行预测。In view of the shortcomings of predicting grade criterion scores with the statistics method, the probability neural networks method was put forward. Computer simulation experiments were conducted under 20 conditions, including the test score as a single variable or multivariable, and different levels of white noise. The results showed that the probability neural networks method could predict grade criterion scores more correctly than the grade regression method under multivariable or white noise conditions.

关 键 词：心理测量 概率神经网络 预测 等级回归 

分 类 号：B842[哲学宗教—基础心理学]