SMS sentiment classification using an evolutionary optimization based fuzzy recurrent neural network.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Classifying spam and ham messages is crucial for effective email management.
• Existing methods can misclassify spam as ham, reducing overall accuracy.
• Sentiment analysis offers a potential avenue to refine message classification.

Purpose of the Study:

• To develop an advanced architecture for accurate spam and ham message classification.
• To enhance classification accuracy by integrating sentiment analysis into the process.
• To introduce a fuzzy-based Recurrent Neural Network optimized with Harris Hawk Optimization (FRNN-HHO).

Main Methods:

• Utilized a Kernel Extreme Learning Machine (KELM) classifier for initial spam and ham message categorization.
• Implemented a fuzzy-based Recurrent Neural Network with Harris Hawk Optimization (FRNN-HHO) for post-classification sentiment analysis.
• Evaluated performance using standard metrics including accuracy, recall, precision, F-measure, RMSE, and MAE.

Main Results:

• The proposed FRNN-HHO architecture demonstrated superior performance in differentiating spam and ham messages.
• Achieved high Area Under the Curve (AUC) values: 0.9699 for SMS, 0.958 for Email, and 0.95 for spam-assassin datasets.
• Sentiment analysis integration significantly improved classification accuracy by addressing misclassifications.

Conclusions:

• The FRNN-HHO model effectively enhances spam and ham classification accuracy through sentiment analysis.
• This approach offers a robust solution for improving text mining and message filtering systems.
• The study validates the effectiveness of FRNN-HHO across diverse datasets, highlighting its practical applicability.