Due to the widespread increase of online reviews, sentiment classification is now a fascinating study in both academic and industrial research. Annotated training data is difficult to obtain, so reviews help in many domains. Numerous methods have been created for sentiment classification, but the process of retrieving information is not exact, less effective, and has a slower rate of convergence. To overcome these problems, this research proposes and evaluates a novel technique for information retrieval and sentiment classification called Dynamically Stabilized Recurrent Neural Network Optimized with Binary Emperor Penguin Optimization Algorithm (SCIR-DSRNN-BEPOA). This approach seeks to enhance the accuracy and efficiency of sentiment analysis by integrating advanced optimization techniques and also to address the challenges such as slow convergence and inefficiency in sentiment analysis, ultimately achieving superior performance metrics. The input data is gathered through Amazon unlocked mobile reviews database and telecom tweets for sentiment categorization with information retrieval. Afterward, the data is fed to the preprocessing. By using the Unscented Trainable Kalman Filter, the preprocessing process removes noise from reviews by fixing typos and removing ellipsis. These pre-processed data is fed to Adaptive and Concise Empirical Wavelet Transform (ACEWT) to extract the features, such as elongated words, punctuation, hash tag, numerical values. The extracted features are given to the DSRNN which classifies the sentiment, like neutral, positive and negative. In general, DSRNN does not show any optimization adaption methods to determine the optimum parameter to offer accurate sentiment classification. Therefore, Binary Emperor Penguin Optimization Algorithm (BEPOA) is proposed to optimize the DSRNN classifier that categorizes the sentiment precisely. Attention Enhanced Temporal Graph Convolutional Network (AETGCN) method is used to extract the necessary review. The proposed SCIR-DSRNN-BEPOA is implemented in MATLAB. To classify the sentiment, the performance metrics, like Precision, Accuracy, F1-score, Recall (Sensitivity), Specificity, Error rate, Computation time, False Alarm Rate, RoC is considered. The performance of the SCIR-DSRNN-BEPOA approach attains 20.11%, 21.12%, 27.73% higher accuracy, 11.13%, 23.04%, 9.51% lesser computation time, 15.29%, 19.43%, 12.45% greater ROC, 28.65%, 23.98%, 27.03% lower false alarm rate when compared with existing methods: Deep Learning improved spider monkey crow optimization algorithm for sentiment analysis and information recovery (SCIR-DRNN), Deep Learning topical level sentiment analysis of social media data (SCIR-LSTM), and a machine learning-based term weighting and feature selection strategy for sentiment analysis of online product evaluations (SCIR-ENN).
Jothilakshmi et al. (Sat,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: