IFM-RCNN: a hybrid text classifier with enhanced performance of binary drug classification from tweets using improved faster mask-recurrent convolutional neural network

• Published in: Knowledge and information systems Vol. 66; no. 1; pp. 557 - 579
• Main Authors: Lavanya, P. M., Sasikala, E.
• Format: Journal Article
• Language: English
• Published: London Springer London 2024; Springer Nature B.V
• Subjects: Artificial neural networks; Classification; Computer Science; Coronaviruses; Customer feedback; Data Mining and Knowledge Discovery; Database Management; Deep learning; Digital media; Drug abuse; Drugs; Health care; Information Storage and Retrieval; Information Systems and Communication Service; Information Systems Applications (incl.Internet); IT in Business; Machine learning; Natural language processing; Neural networks; Performance enhancement; Regular Paper; Social networks; Viral diseases; Drugs; Clinical word extraction; Performance measures; Social media; Natural-language-processing
• ISSN: 0219-1377; 0219-3116
• DOI: 10.1007/s10115-023-01957-9

The significant growth of social media and online websites helps to garner a significant amount of healthcare data and contributes toward the development of the healthcare industry. Authorized health assets, which include specialized healthcare authorities, medical professionals, substandard monitoring and diagnosing equipment, and pharmaceuticals, became insufficient during the Coronavirus Disease (COVID-19) pandemic in 2019. Social networks have played a significant role in selling drugs during those times. However, the risk of drug abuse is increased, and also there occurs several challenges in the management and detection of drug activities. This research will focus on identifying relevant features in the drug database by developing a new target label model using deep learning approaches. This helps in avoiding unbalanced data, which in turn results in enhanced overall performance. The objective of this work is to address uneven classification performance by using social media comments as input in order to improve the accuracy of the training model. A new powerful hybrid machine learning method to extract clinical terms from customer feedback and learn the system by categorizing medical conditions and drug names is developed in this work. The Improved Faster Mask-Recurrent Convolutional Neural Network (IFM-RCNN), along with natural language processing, is trained with 4589 hand-labeled examples with 18,465 synthetically generated tweets. The suggested IFM-RCNN model attained an (area under the ROC curve) AUC of 0.98 and 0.94, a precision of 86.95% and an accuracy of 92.65%, respectively. The results show an enhanced capacity to anticipate drug activities with minimal differentiation. The proposed methodology has improved performance. It can be utilized for clinical applications, and the findings revealed that the suggested model can be applied in real-time applications.