Method Research For The Prediction Of Drug Side Effects Based On Recommendation Algorithm

Drugs may be accompanied by side effects that endanger the patient's health while treating the patient's illness.Therefore,how to quickly and accurately discover potential side effects becomes a vital part of the drug development process.Traditional methods of discovering drug side effects are usually through biology and pharmacology experiments,which tend to take a long time and huge capital investment.Therefore,it is important to accurately and quickly predict the potential side effects of drugs through computer simulations.At present,most methods of drug side-effect prediction use a machine learning classifier to predict side-effect categories.However,due to the sparseness of the matrix of drug side effects,the prediction effect is often greatly affected.At the same time,the traditional drug side-effect prediction methods usually only pay attention to the information of the drug itself,and ignore the potential correlation between side effects.In view of these,this paper proposes a side-effect prediction method based on the recommendation algorithm.The method takes drugs and side effects as the research object,the drug side effect prediction problem is regarded as a recommendation system,and the recommended algorithm is used to combine the relationship between the drug information and drug side effects to recommend possible drug side-effect relationships,and it provides a reference for drug experiments.The main contents are as follows:1.According to the potential correlation among side effects,we selected the side effects related to central nervous system as the research object,and proposed a prediction method based on collaborative filtering algorithm.The method uses a collaborative filtering algorithm to add drug penetration information in the blood-brain barrier to predict drug side effects.The experimental results show that the prediction method based on collaborative filtering algorithm can effectively improve the prediction accuracy of central nervous system-related drug side effects.2.Taking the existing drug side effects as the research object,we proposed a method for predicting drug side effects based on matrix reconstruction and social matrix decomposition.This method mainly uses a user-based collaborative filtering algorithm to mine hidden information from the correlation of drug side-effects and reconstruct the sparse drug side-effects correlation matrix.At the same time,it uses an improved multifeature fusion social matrix decomposition method to combine multiple drug information to mine potential association between side effects,and it improves the accuracy of drug side effects prediction.Through comparison experiments on multiple data sets,the result verified the effectiveness of these two methods,the matrix reconstruction method based on collaborative filtering and the social matrix decomposition method.3.We developed a tool of drug side-effect prediction based on recommendation algorithm.The tool combines multiple drug-related features to predict drug side effects.While providing prediction results of drug side effects,it is convenient for users to understand the degree of correlation between drug side-effect prediction and various drug information.Therefore,it has a certain reference value.