Classification of People’s Opinions on Fuel Subsidy Removal in Nigeria: An Enhancement of Unsupervised Learning Algorithm-Opinion Lexicon algorithm

Abstract

This research aims to  collect  Nigerians’  opinions  on  subsidy  removal  in  Nigeria  and  classify  them  using  an  unsupervised  learning  algorithm,  specifically  corpus-based  lexicon  algorithm  in  order  to  enhance  it  prediction  accuracy.  The  data  was  extracted  from  an  online  survey  via  social  media  platforms:  Facebook  and  WhatsApp.  A  comprehensive  literature  review  has  been  conducted  on  fuel  subsidy  removal,  sentiment  analysis,  and  unsupervised  machine  learning  approach.  The  methodology  involved  are:  data  collection,  preprocessing,  feature  extraction,  model  training,  and  evaluation.  The  result  of  this  study  shows  that  Nigerians  are  not  happy  with  fuel  subsidy  removal,  because  of  the  highest  number  of  negative  comments  over  positive  ones.  This  implies  that  there  exist  in  socio-economic  and  security  issues  in  Nigeria.  The  unsupervised  learning  algorithm  for  sentiment  analysis,  the  Lexicon  was  improved  with  an  accuracy  of  84.5%. thus, enhanced by 15.27%. These results can potentially inform policymakers and stakeholders about the public's sentiments, the social and security consequences of subsidy removal in Nigeria.