Sentiment Analysis using Hybrid Approach on Patients Feedback in Context of Multispecialty Hospitals in Raipur

Abstract

newline This thesis presents a comprehensive study on Sentiment Analysis of Patient Feedback in the Context of Multispecialty Hospitals in Raipur; apply a novel Stacking Hybrid Approach to improve predictive accuracy and reliability. The growing reliance on patient feedback for assessing healthcare quality has necessitated the development of tough analytical tools capable of deciphering intricate sentiments embedded in patient reviews. This research proposes the Hybrid Stacked Sentiment Analyzer (HyStSA) model, which is specifically considered to outperform conventional individual sentiment analysis models by leveraging the strengths of multiple machine learning techniques. newlineThe application of Fine-Grained Sentiment Analysis (FGSA) could significantly enhance the depth and precision for analysis. FGSA is designed to identify and categorize sentiments at a more nuanced level, distinguishing between various degrees of positivity and negativity, such as quotVery Satisfied,quot quotSatisfied,quot quotNeutral,quot quotDissatisfied,quot and quotVery Dissatisfied.quot This aligns perfectly with the structured feedback form, which collects patient opinions across multiple categories. newlineEssential to this research is a novel binary conversion method known as Binarizing Categorical Variables (BCV). This method extensively improves model accuracy by transforming categorical data into a binary format that improved aligns with the requirements of machine learning algorithms. Moreover, a correlation matrix is employed for feature selection, ensuring that only the mainly applicable features are utilized in the model, further enhancing its performance. The effectiveness of this approach is established through the analysis of patient feedback from multispecialty hospitals, where the sentiment data is often complicated and versatile. newlineThe HyStSA model is constructed by combining numerous base classifiers Logistic Regression (LR), Naive Bayes (NB), Random Forest (RF), Gradient Boosting (GB), and Support Vector Machine (SVM) into a stacked ensemble structure.