Prediction method of prescription efficacy based on multi-layer perceptron and natural language processing technology

Abstract

The invention discloses a prescription efficacy prediction method based on a multilayer perceptron and a natural language processing technology, which comprises the following steps: collecting relatedtraditional Chinese medicine data, and integrating the data to form a prescription library and a disease library; Constructing a traditional Chinese medicine field basic corpus; Establishing a medical effect language model by utilizing a natural language processing technology; Obtaining a prescription main treatment mark matrix by improving a word vector weighting method; Constructing a prescription composition matrix according to the prescription medicinal material composition and the medicinal material dosage; Constructing a training set according to the prescription composition matrix andthe information of the prescription main treatment mark matrix; Based on the training set, performing training by using a multi-layer perceptron, and establishing a prediction model; And performing efficacy prediction on the to-be-tested prescription based on the prediction model. According to the method, the natural language processing technology is used for automatically marking the prescriptioncharacteristics, the workload is greatly reduced, the characteristics of the multi-layer perceptron are used for modeling a large amount of traditional Chinese medicine prescription data, the data are output in a probability mode, and accurate and visual judgment of the reliability is facilitated.

Description

technical field [0001] The invention relates to the field of computer-aided drug design, in particular to a prescription efficacy prediction method based on multi-layer perceptron and natural language processing technology. Background technique [0002] New drug research and development is a huge systematic project. The successful development of a new drug costs an average of more than 1 billion US dollars and takes more than 10 years. However, in the past 20 years, with the increase in research and development investment year by year, the output of new drugs (that is, the number of new drugs on the market) has shown a downward trend. [0003] It is generally believed that the traditional "single drug-single target" drug development model is an important reason for this dilemma. In fact, most diseases, especially complex diseases, are difficult to be effectively treated with a single drug and a single target. Due to the development of related technologies such as life scie...

AI Technical Summary

Problems solved by technology

However, the use of these traditional machine learning methods faces many problems: First, the large number of traditional Chinese medicine prescriptions makes it difficult to realize the huge workload of manual labeling; second, there is no accurate indicator to judge whether the prediction results are correct; and when the amount of data is huge, the The processing of traditional Chinese medicine data has also become a big problem

Images