Synthetic prediction of MOF based on RF and K-nearest neighbor

In this paper, we propose a MOF synthetic prediction model that integrates the Random Forest (RF) and K-nearest neighbor (KNN) algorithms. By applying advanced machine learning techniques to analyze MOF synthesis data, the model learns from historical synthesis records to automatically predict the synthesis conditions and properties of new MOF materials. The RF algorithm employs an ensemble approach, utilizing multiple decision trees to effectively capture complex nonlinear relationships and reduce overfitting, enhancing model robustness. Meanwhile, KNN offers a non-parametric method that excels in local approximation, ensuring precise predictions based on the nearest historical data points. By combining these algorithms, the model leverages RF's capability to handle high-dimensional data and KNN's sensitivity to local variations, resulting in superior prediction accuracy and stability. Moreover, feature importance analysis from the RF component provides insights into key factors influencing MOF properties, facilitating informed decision-making in the synthesis process. Experimental findings confirm that the model not only achieves high accuracy but also demonstrates strong generalization in predicting MOF synthesis.