Chemical reaction validity predicting method

Abstract

The invention discloses a chemical reaction validity predicting method. The method comprises the specific steps of segmentation, specifically, the chemical reaction is subjected to information acquisition, the elements in the chemical equation are segmented through a machine learning module, and independent elements are obtained; vectorization, specifically, the elements after being segmented serve as vocabulary units to built an element table, the elements are mapped into the vector space of specific dimension through a word embedding method till all elements are represented by vectors; result prediction, specifically, compound feature representation and chemical reaction internal feature representation are conducted by the machine learning module, the feature space representing the result is enabled to pass through the fully connected layer to acquire feature representation of an initial compound and a target compound, validity judgment is conducted by calculating the distance between vectors, and the success rate prediction result is acquired and output; and sorting and updating, specifically, the history data of the chemical reaction is divided by the machine learning module into a positive example and a negative example according to success or failure, the system parameters are updated in combination with loss function and penalty term, so that the result tends to be moreand more precise.

Description

technical field [0001] The invention belongs to the intersecting field of computer science and chemical organic synthesis, and relates to a method for feature extraction and analysis of chemical reactions based on machine learning technology, for verifying the legality of organic synthesis, and for predicting the success rate of unknown chemical reactions. Background technique [0002] Machine learning shines in many fields such as biopharmaceuticals and medical diagnosis. It has changed traditional research methods, improved scientific research efficiency, and prompted changes in many industries. By learning the deep information hidden in the data and mining the internal correlations to make predictions and judgments, the machine learning system has extremely effective insights and high efficiency, and can achieve capabilities comparable to or even surpassing humans in vertical fields. [0003] The traditional organic synthesis mode is not only a challenge to the cost and p...

AI Technical Summary

Problems solved by technology

[0003] The traditional organic synthesis mode is not only a challenge to the cost and physical strength of experimental workers, but also a challenge to the memory and analysis ability of route designers

Images