Using deep learning methods for discovering associations between drugs and side effects based on topic modeling in social network
| Autor: | Hossein Ebrahimpour-Komleh, Majid Fouladian, Zahra Rezaei, Behnaz Eslami, Mehdi Habibzadeh |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Topic model
Drug Text corpus Computer science Association (object-oriented programming) media_common.quotation_subject 02 engineering and technology 03 medical and health sciences Side effect (computer science) 0202 electrical engineering electronic engineering information engineering Media Technology 030304 developmental biology media_common 0303 health sciences Information retrieval business.industry Communication Deep learning Computer Science Applications Human-Computer Interaction Ask price 020201 artificial intelligence & image processing Artificial intelligence business Classifier (UML) Information Systems |
| Zdroj: | Social Network Analysis and Mining. 10 |
| ISSN: | 1869-5469 1869-5450 |
| DOI: | 10.1007/s13278-020-00645-8 |
| Popis: | The relationship between drug and its side effects has been delineated in two websites, namely Sider and WebMD. The aim of the present paper is to find the relationship between drug and its side effects as reported by typical users of a website called Ask a patient, and to compare these reports with the side effects in reference sites. In addition, the typical users’ comments on highly-commented drugs (neurotic drugs, anti-pregnancy drugs and digestion drugs) within last decade were analyzed. The reason for such investigation is the fact that typical users’ comments and their tendencies can be considered as an important factor in determining the best drugs in improving them or decreasing their risk dangerous. Typical users’ comments on drugs’ side effects were gathered from the website Ask a patient. Then, the data on drugs (neurotic drugs, anti-pregnancy drugs and digestion drugs) were classified according to deep learning model. At first using the model, the three issues, namely drug, its side effect and the cause of the side effect, were explained. Afterward, using topic modeling, the main topics of side effects for each group of drugs were identified. Finally, using the websites of Sider and WebMD in which the side effects of drugs are reported, the side effects of the three classes of drugs were retrieved. The goal of the present research was to analyze typical users’ comments reported on the website called Ask a patient, and to compare these comments with the reports about the side effects of drugs from important sites. Our model demonstrates its ability to accurately describe and label side effects in a temporal text corpus. By taking full advantage of deep learning classifiers, the used methods in text mining is shown to be accurate and effective for discovering association between drugs and side effects. Moreover, through combining with modular classifier in addition to topic modeling, this model has the capability to immediately locate information in reference sites to recognize the side effect of new drugs. In fact, due to the unbiased nature of typical users’ comments, these comments can be a reliable indicator for drug producer companies to reduce the side effects of drugs. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink