Research on the Prediction of Recycled Thermal Insulation Concrete Strength by Optimizing Different Machine Learning Models

As a new type of green concrete, recycled thermal insulation concrete is of great significance to the development of green circular economy and efficient energy-saving buildings. Combining sparrow search algorithm (SSA) with random forest (RF) and Elman neural network principle, a SSA-RF model with higher prediction accuracy is designed. Based on the data obtained from the strength experiment of recycled thermal insulation concrete, a network model for predicting the strength of recycled thermal insulation concrete is established. Using the replacement ratio of recycled coarse aggregate, water-cement ratio and concrete apparent density as factors, a rapid prediction model for the strength of recycled thermal insulation concrete is established based on random forest (RF), Elman neural network, sparrow search optimization random forest (SSA-RF) and sparrow search optimization Elman (SSA-Elman), respectively. Select 61 groups of experimental data reported in the literature and carry out case study, the results show that: The SSA-RF model performs best, with high prediction accuracy for the strength of recycled thermal insulation concrete, followed by SSA-Elman, and the prediction accuracy of unoptimized RF and Elman neural network is low.