Survey and Evaluation of Extreme Learning Machine on TF-IDF Feature for Sentiment Analysis

EasyChair Preprint no. 9421

Abstract

Sentiment analysis is a tool to understand the emotion of the statement given by customers. This understanding helps the service provider to in improving the quality of service. Machine learning models are one of the popular choices for designing sentiment analysis systems. However, hyper-parameter tuning is one of the important concerns in most of these models. Moreover, the gradient-based training models are prone to the local-minima problem. In such a case one-pass learning model like Extreme Learning Machine (ELM), gives generalization performance with minimal hyper-parameter tunning. This work studies in depth the ability of ELM to learn a generalization model for the sentiment analysis problem. Here the study uses the airline twitter review dataset to empirically analyze the ELM model and the required hyper-parameter setting.

Keyphrases: ELM, machine learning, Sentiment Analysis, TF-IDF

@Booklet{EasyChair:9421,
  author = {Manpreet Kaur and Dibyasundar Das and Smita Prava Mishra},
  title = {Survey and Evaluation of Extreme Learning Machine on TF-IDF Feature for Sentiment Analysis},
  howpublished = {EasyChair Preprint no. 9421},

  year = {EasyChair, 2022}}