The rational synthetic parameter analysis for subclasses of microporous aluminophosphates based on hierarchical feature selection model

• A Hierarchical Feature Selection Model is proposed to analyze the rational synthetic parameter of the AlPOs subclass.
• A method called SDAS is proposed to measure the weight of the features for data containing several subclasses.
• The synthetic parameters of AlPOs are analyzed by considering the subclass and the correlation of features simultaneously.

Open-framework aluminophosphates (AlPOs) is an important family of porous crystal materials. But the synthetic chemistry of this kind of materials is very complicated, and the synthesis mechanism has not been clearly understood yet. In this paper, we propose a Hierarchical Feature Selection Model (HFSM) composed of two layers to analyze the rational synthetic parameters for the subclass of microporous aluminophosphates (AlPOs) containing (6,8)-rings. In the first layer, we select a feature subset that could separate the (6,8)-ring-containing microporous AlPOs from other AlPOs. In the second layer, we further analyze which of these selected features are critical for the formation of each special subclass in (6,8)-ring-containing microporous AlPOs. With the optimal feature subset selected by the proposed model, we can obtain the highest accuracy rates as 94.28%, 94.03%, 91.27% and 92.20% for the classification of AEN, AWO, CHA and ERI, respectively. Extensive analysis is presented for the synthetic parameters selected by the hierarchical model, which could provide a useful guidance to the rational synthesis of such materials.

Figure optionsDownload as PowerPoint slide