Data mining with 12 machine learning algorithms for predict costs and carbon dioxide emission in integrated energy-water optimization model in buildings

In recent years, various models, which employ green and sustainable energy-supplying systems, have been presented in different ways to optimize water and energy consumption in several countries, pursuing the purpose of reducing water and energy consumption and costs. The present study considered an integrated and unconcentrated water and energy consumption optimization model in the building and investigated it economically and environmentally using mixed-integer linear programming. In this research, the data of different sections, including the data related to the climatic conditions, environment, costs, technologies, etc., were collected. Next, the researchers sought to predict the model results using 12 highly accurate machine learning algorithms and considering four indices to examine the prediction accuracy of the algorithms. The utilization of the machine learning prediction algorithms helped us economically and environmentally investigate and predict the model conditions in every geographical location with respect to the data that varied according to the climatic conditions. The results obtained from investigating the indices of the algorithms for the data examined for the objective functions, including cost optimization and carbon emission reduction, revealed that the prediction accuracy ranged from 0.8 to 0.96, and 0.79 to 0.91 for the first and second objective functions, respectively, in the 12 examined algorithms. Meanwhile, the Light Gradient Boosting Machine and Extra Tree algorithms enjoyed higher prediction accuracy in this research than other algorithms. Next, to analyze the results, the researchers implemented the Principal Component Analysis method to reduce the input data dimension to the algorithms. The results of the algorithms reflected a decline in the prediction accuracy after the dimension reduction. Finally, the effective variables in the prediction accuracy of the algorithms were presented by using the Stepwise Regression method for both objective functions. Overall, the output results of the algorithms showed high prediction accuracy for investigating the model conditions in various geographical regions.

Automated summary

This study optimized the unconcentrated water and energy consumption in buildings using mixed-integer linear programming and investigated it economically and environmentally with 12 machine learning algorithms, with results showing higher prediction accuracy in Light Gradient Boosting Machine and Extra Tree algorithms.