Technical Retrofitting Project Cost Prediction Based on Least Squares Support Vector Machine

Predicting the cost of equipment technical retrofitting projects can provide crucial support for annual investment decisions in power grids. Considering the high dimensionality of the features and the nonlinear correlations among them, an improved least squares support vector machine is developed. First, the input data are processed using principal component analysis to reduce data redundancy. Then, a sparrow search algorithm is employed to optimize the regularization parameters and kernel function parameters, which advances in seaching the best parameters of the support vector machine. At last, the processed data are imported into the constructed model for training and prediction. Compared with the traditional support vector machine and backpropagation neural network model, the proposed method exhibits better generalization capabilities and higher prediction accuracy.