An IBGWO optimized feature selection framework for sentiment analysis-based fraudulent website detection using MLRNN

Description

Saraswathi Perumal and Anchitaalagammai Jayalakshmi Vishwanathan present a 2026 method for spotting fraudulent websites with machine learning. Scam sites threaten users, and older detectors struggle with large feature sets, redundant signals, and new attack styles. Their pipeline starts from a Kaggle dataset of web page phishing examples holding 11,430 URLs described by 89 features. A first pass trims weak features in rounds, and a tuned grey wolf search then picks a compact subset, judging each candidate subset with a fitness function.

The chosen features feed a multi-layer neural network that classifies pages, drawing partly on sentiment-style signals in page content. Performance was checked by rotating training and test portions ten times. The authors report precision, recall, accuracy, error rate, and F-measure from the confusion matrix, and state that the combined method outperformed earlier approaches on this dataset.

Catalog record

Source: https://europepmc.org/article/MED/PMC13500837