The traditional class ratio compression transformations is only fit for the monotone sequence. This paper puts forward the criterion which makes the fitting precision not higher than its inverse fitting precision, when the fitting values are inversed. Then, a new criterion is given of the class ratio dispersion of compression transformations. At last, a new transformtion is gotten which makes the inverse fitting precision be improved.%传统的级比压缩变换仅适用于单调序列,针对单调序列,给出模型预测值还原后,还原精度不变或者大于其建模精度的数据变换的判定定理,提出了新的级比偏差压缩变换的构造准则.最后构造出还原精度不变或者提高的级比偏差压缩变换,从理论上验证了该判别法的实用性和有效性.
展开▼