An Empirical Study of the Extreme Learning Machine for Twitter Sentiment Analysis
International Journal of Intelligent Systems and Applications in Engineering
Extreme Learning Machine (ELM) method is proposed for single hidden layer feed-forward networks (SLFNs). The ELM employs feed-forward neural network architecture and works with randomly determined input weights. In this respect, ELM depends on the principle that enables to determine weights and biases in the network. In the first phase of ELM which can be named as feature mapping, random values are used differently from other methods such as Support Vector Machines (SVM) and Deep Neural
... Deep Neural Networks that employ a kernel function for this purpose. After the feature mapping step, the main goal of the ELM is to learn weights between hidden and output layers by minimizing the error. Therefore, the ELM has gained much more popularity recently; and can be applied for solving various problems like classification, regression, and dimension reduction. In this study, our aim is to apply the basic ELM for making sentiment analysis from Twitter messages as it is considered as a classification task in the literature. To evaluate the performance of the ELM for sentiment analysis, we compare it with the SVM which is one of the most successful machine learning algorithms used for sentiment analysis. To our knowledge, this paper is the first study that employs the ELM for sentiment analysis from Twitter messages written in Turkish. Experimental evaluation of the two methods are done by using two different Turkish Twitter messages datasets. The experimental results showed that the performances of the two methods are slightly different, however SVM outperforms basic ELM.