A Bidirectional Encoder Representations Transformer based Pharmaceutical Survey Data Collection and Analysis System

The science and industry of developing, producing, and marketing medicinal drugs, vaccines, and other healthcare products are referred as Pharmaceutical. The existing methods of Pharmaceutical survey data collection and analysis has issues data quality issues, survey fatigue, and insufficient consideration of variables. To overcome these problems, a BERT model is proposed, where a Unified Theory of Acceptance and Use of Technology 2 (UTAUT 2) data set is considered which evaluates digital health technologies, and enhances data quality. This dataset is given as input for preprocessing, which is performed using K-Means Clustering (KMC) to improve data understanding and quality. This is followed by data validation, to identify behavioral intention of model using Structural Equation Modeling (SEM). Next, Time Series Analysis (TSA) which is performed using Seasonal Autoregressive Integrated Moving Average (SARIMA) to handle seasonality and trends in survey data. For pretraining, Robustly Optimized BERT (ROBERT) approach is taken to analyze sentiment, and answer questions. Also, Bidirectional Encoder Representations Transformer (BERT) is used for analysis to improve accuracy, and to handle complex medical texts. The proposed BERT model attained better accuracy, precision, recall, and f1 score of 96.87%,95.72%,93.45%,97. 24%, as compared to the existing approaches like K-Nearest Neighbor (KNN).