In addition , in order to guarantee the smoothness of weight vector distribution in output space , mfknn2 method is introduced a smooth factor in the modification of connective weight coefficient between neuron 同時(shí)對mfknn2中神經(jīng)網(wǎng)絡(luò)的權(quán)值修改還引進(jìn)一個(gè)平滑因子,以保證權(quán)矢量在輸出空間分布的平滑性。
We design a model based on bp neural network with a smoothing factor to predict the market share of the enterprise product . the result of experiment shows that the forecasting precision of this model is higher than the classical model 本文提出了利用平滑bp神經(jīng)網(wǎng)絡(luò)模型進(jìn)行商品市場占有率時(shí)序預(yù)測的方案,并通過實(shí)驗(yàn)證明:這種模型比一般的bp神經(jīng)網(wǎng)絡(luò)模型預(yù)測精度稍高。
The result of experiment shows that this scheme is feasible . a bp algorithm with smoothing factor is proposed to optimize the weighting space . the result of experiment shows that this scheme is useful for improving the bp neural network ' s generalization ability 對于bp神經(jīng)網(wǎng)絡(luò)的優(yōu)化,本文還提出了一種帶平滑因子的bp算法,通過在bp算法中嵌入平滑因子,對權(quán)值空間進(jìn)行平滑優(yōu)化,并通過實(shí)驗(yàn)證明這種方法有助于bp網(wǎng)絡(luò)性能的提高。