Sentiment analysis via multi-layer perceptron trained by meta-heuristic optimisation
Dabiah Alboaneen, Huaglory Tianfield, Yan Zhang · 2017
In this paper, a new tweet analysing approach is proposed, which is composed of two main phases; feature selection and tweets classification. In the first phase, mutual information (MI) is used to select the best set of features to reduce the feature dimensions. In the second phase, a meta-heuristic algorithm is used to optimise weights and biases of multi-layer perceptrons (MLPs) network and then implemented to classify twitter sentiments. Experimental results on existing twitter dataset show better performance of the glowworm swarm optimisation (GSO) based MLP over genetic algorithm (GA) and biogeography-based optimisation (BBO) algorithms.