Design And Implementation Of Medicine Recommendation System Based On Text Mining And Heterogeneous Information Network

With the rapid growth of medical literature data,the research of medical big data are in full swing.Recently,how to tap effective medical knowledge from huge literature data for clinical decision support of doctors is also a hot topic.At present,the drug recommendation algorithm is an important solution to this research.Drug recommendation algorithm can be divided into drug-based recommendation algorithm and disease-based recommendation algorithm.In the drug-based recommendation algorithm,a method of extracting drug interaction relationships based on text mining is proposed for problem that insufficient feature extraction,unscientific feature processing,etc.In the recommendation algorithm based diseases,mining whether there is a relationship between the drug combination,put forward a combination drug based on heterogeneous network information recommendation algorithm.The main tasks of the study include:(1)extract the drug-drug interaction based on text mining.This study has studied and improved the extraction methods of existing drug interaction relationships.This paper finds that the original method has many deficiencies in the selection and processing of features.To address these deficiencies,this study proposes a method for extracting drug interactions based on text mining.This method requires extraction of features,processing characteristics,and classification.In the feature extraction section,features such as keywords are added;in the feature processing section,word vectors are used to quantify the features;in the classification section,the classification model in machine learning is used for experiments.Experiments show that the performance of the extraction method in this paper is superior to other methods in the same data set.(2)Combinatorial drug recommendation algorithm Com-MedRank based on medical literature.This study has studied and improved the existing HIC-MedRank algorithm that finds effective antihypertensive drugs from medical literature library MEDLINE.This study improved the relationship between drugs without considering the original algorithm and extracted the drug interaction relationship.The model is applied to the algorithm to judge whether the relationship between the drugs is a combination relationship,and the input of the algorithm is modified.The experimental input of this study uses the hypertension,the experimental output is the antihypertensive combination drug,and this study recommends the combination drug for the clinician.(3)The implementation of medical intelligence recommendation system.The system is built based on the framework technology of Java and JUNG tools.The system is mainly divided into two modules: drug interaction relationship extraction module and drug recommendation module.In the drug interaction relationship extraction module,this function can obtain drugs which can generate interaction and their interactions type;In the drug recommendation module,the user obtains a list of drugs whose Top 10 disease is based on the disease keyword.In both modules,the system uses the JUNG toolkit to visualize the network graph and visualizes the generated network through the relational network diagram.The system has passed all test cases to meet the on-line requirements?...