A novel approach based on an extended cuckoo search algorithm for the classification of tweets which contain Emoticon and Emoji

Twitter is a popular microblogging service that allows its users to view and share limited character messages (known as “tweets”) with the public. This paper proposes a tweet sentiment classification framework which pre-processes information from Emoticon and Emoji in such way that their textual representation is included to enrich the tweet. Once the tweets are pre-processed, a hybrid computational intelligence approach is applied for classifying the tweets into positive and negative. The proposed hybrid approach is based on a compilation of several methods: the Singular Value Decomposition and dimensionality reduction method to reduce the dimensionality of the data; the Extended Binary Cuckoo Search algorithm, to further reduce the matrix by selecting the most suitable dimensions; and the Support Vector Machine classifier which is trained to identify whether a tweet is positive or negative. The experimental results revealed that the proposed approach yields higher classification accuracy and faster processing times than the baseline model which involves applying the Extended Binary Cuckoo Search algorithm and Support Vector Machine classifier to the original matrix, without using Singular Value Decomposition and dimensionality reduction.