Predicting the AC Conductivity of Nanocomposite Films using the Bagging Model

As a high-endurance, high-temperature insulating material, the insulating property of polyimide must be improved, and the optimal synthesis of polyimide-based composite needs to be further explored. In this paper, the AC conductivity of multiple series of polyimide filled nanocomposite films are simulated by the Bagging Model. The prediction results show that the multi-series of polyimide-based nanocomposite films, which are filled with different inorganic nanoparticles and different doping contents, are predicted by 5-fold cross validation. The predicted value of the model fits well with the practical measured value. Further analysis of the results shows that the mean absolute error, mean squared error, and root mean squared error of the Bagging Model are less than those of the Linear Regression, Medium Tree, Coarse Tree, support vector regression (SVR), Gaussian process regression of the RQ nucleus, K-nearest neighbor, Random Forest, and AdaBoost models. Thus, for predicting the AC conductivity of polyimide nanocomposite films, the Bagging Model is better than other models to provide theoretical direction for the design of a high insulation polyimide composite. This modeling method may effectively reduce the time required for researchers to design nanocomposite films, greatly improving their efficiency and reducing the cost of research in the future.