Using Bio-inspired intelligence for Web opinion Mining

This work proposes a bio-inspired based methodology in order to extract and evaluate user's web texts / posts. To validate the methodology, a dataset is constructed using real data arising from Greek fora. The obtained results are compared with a commonly used machine learning technique (decision treesC4. 5 algorithm). The bio-inspired algorithm (namely the hybrid PSO/ACO2 algorithm) achieved average classification accuracy 90. 59% in a 10 fold cross validation experiment, outperforming the C4. 5 algorithm (83. 66%). The proposed methodology could be easily integrated with a decision support system providing services in the fields of e-commerce or e-government in order to help merchants acquire customer satisfaction or public administrators capture common understanding.