A Classification Model with Cognitive Reasoning Ability

In this paper, we study the classification problem of large data with many features and strong feature dependencies. This type has shortcomings when handled by machine learning models. Therefore, a model cognitive reasoning ability is proposed. The core idea to use mechanism proposed in paper solve structured multiple correlation between features, then implements for features. three parts. first part proposes Feature-to-Image algorithm converting into image data. quantifies dependencies so as take account impact individual independent correlations on prediction results. second designs low-level extraction quantified using convolutional neural networks. With relative symmetry capsule network, third implement high-level extraction, reasoning, tasks At same time, provides derivation process description mechanism. Experiments show that our efficient outperforms comparable models category experiment ADMET properties five compounds.This work will provide new way computing intelligent analysis.