Twitter Sentiment Analysis via Chaotic Quantum Fruit Fly Optimization: Enhancing Feature Selection and Classification
Palanisamy SatheeshKumar, Jeevitha Kandasamy, Nivethitha Thangavelsamy, Rachana Arya, Nageswari Dhandapani, Sghaier Guizani · Mathematical Modelling and Engineering Problems · 2025
This research presents a novel feature selection framework-Chaotic Quantum Fruit Fly Optimization Algorithm (CQFOA)-designed to enhance Twitter sentiment analysis.CQFOA extends the standard Fruit Fly Optimization Algorithm (FOA) by incorporating two advanced mechanisms: (1) a chaotic mapping strategy to maintain population diversity and prevent premature convergence; and (2) quantum-behaved position updating using probabilistic rotation gates for global exploration.The integration of these strategies improves the algorithm's ability to handle the high dimensionality and the noise characteristic of Twitter data.CQFOA was applied as a feature selector prior to classification by Convolutional Neural Networks (CNN); Recurrent Neural Networks (RNN) and Recursive Neural Networks.Compared to traditional feature selection techniques-Particle Swarm Optimization (PSO); Genetic Algorithm (GA); Artificial Bee Colony (ABC); and the conventional FOA-CQFOA achieved higher performance across multiple evaluation metrics.In particular, average improvements were observed in accuracy (up to 94.13%); precision (96.84%); recall (97.24%); and F-measure (98.97%).These results were validated using 10-fold crossvalidation and assessed via paired t-tests, confirming statistically significant improvements (p < 0.05) over baseline methods.To ensure reliability, class distribution and data preprocessing strategies were rigorously monitored to mitigate overfitting and class imbalance.The proposed CQFOA framework demonstrates robustness in highdimensional noisy data environments and offers a reproducible pipeline for sentiment classification tasks in social media analytics.