Improving Energy Efficient Aspect of Spam Classification Framework using Ensemble Machine Learning

Authors

  • S.Selva Birunda, Dr. R. Kanniga Devi

Abstract

The majority of Machine Learning research focuses on obtaining accuracy at a higher level. Evaluating the Machine Learning model in terms of energy consumption is emerging. This work serves as an insight to reduce carbon emission and energy in spam classification framework. An unsolicited text message, called Spam created various threats to the users. However, identifying spam messages and filtering them is one of the challenging tasks. The objective of this paper is to analyze the predictive performance of different supervised machine learning classifiers to classify spam messages and making the Machine Learning models to consume less energy using Energy Efficient Voting Ensemble algorithm. Earlier, no other work considered energy efficiency in Spam Classification. Initially, the dataset undergoes preprocessing, which is followed by extracting the relevant features from the dataset using the Bag of Words and the Term Frequency-Inverted Document Frequency technique. The processed dataset undergoes training using multiple supervised Machine Learning classifiers along with the Voting Ensemble technique, and their predictive performances are compared based on accuracy, precision, recall, and F1 score. Besides, we consider energy efficiency as a metric for reducing energy consumption would be better to reduce carbon footprint and hence make it environment friendly.

Keywords-supervised machine learning; ensemble; spam; filtering; energy efficiency; carbon footprint

Published

2020-12-31

Issue

Section

Articles