基于PyMC的线性混合模型及其在印染工业上的应用

The Linear Mixed Models Based on PyMC and Its Application in Printing and Dyeing Industry

理论上通过线性模型,只用染料的浓度配比就可以预测染出何种颜色。但实验表明面料也会影响染色结果。因此本文将面料作为随机效应加入到模型中,构造线性混合模型。本文用机器学习库Py MC实现线性混合模型,并用该模型改进了基于线性回归的预测模型。实验表明线性混合模型优于普通线性回归,同时证明了面料确实会影响染色效果。根据模型的随机效应,还可对面料进行分类。

Theoretically, it is possible to predict the color by using only the dye concentration ratio by linear models. But experiments show that fabric can also affect dyeing results. Therefore, fabric is added into the model as a random effect to construct a linear mixed model. In this paper, PyMC, a machine learning library, is used to implement the linear mixed model and improve the prediction model based on linear regression. The results show that the linear mixed model is superior to the ordinary linear regression, and the fabric can affect the dyeing effect. Fabrics can also be classified according to the random effects of the model.