Table 1 Summary of related work carried out in sentiment based classification.
Study & Year | Key highlights | Advantages | Disadvantages |
|---|---|---|---|
Pak and Paroubek11 | Used graph extraction for document classification; documents represented as trees with interconnected nodes | Provides structured representation of document dependencies | Potential loss of fundamental emotions in text representation |
Rafrafi et al.12 | Employed neural network AI model for sentiment classification using the Amazon corpus | Achieved high accuracy using deep learning models | High computational complexity compared to traditional classifiers |
Zhang et al.13 | Classified Chinese firm smartphone reviews using SVM and Naive Bayes with a numeric sentiment scale | Effectively categorized sentiment using traditional classifiers with numeric scoring | Limited adaptability to highly unstructured textual data |
Xue et al.21 | Introduced self-adaptive PSO (SaPSO) for large-scale feature selection | Improved feature selection for large datasets, reducing dimensionality issues | Limited adaptability to dynamic and evolving datasets |
Xue et al.22 | Used Non-Dominated Sorting Algorithm-III for feature selection with missing data | Handled feature selection in datasets with missing data, improving robustness | Performance depends on the quality of missing data handling |
Xue and Qin23 | Proposed ADARTS, leveraging channel attention for feature selection to enhance efficiency and memory use | Enhanced search efficiency and memory utilization through channel attention mechanisms | Requires extensive computation for feature selection optimization |
Suddle et al.24 | Optimized LSTM architecture using meta-heuristics for sentiment analysis | Achieved optimized LSTM models for improved sentiment classification performance | Meta-heuristics tuning is complex and computationally expensive |
Dangi et al.25 | Developed an improved Robust Random Vector Functional Link Network (RRVFLN) using Artificial Rabbits, incorporating FastText for word vector analysis | Enhanced classification accuracy through improved feature selection and token-level analysis | Requires extensive preprocessing and fine-tuning for optimal performance |
