Review-based analysis of clustering approaches in a recommendation system

Abstract

Because  of  the  explosion  in  data,  it  is  now  incredibly  difficult  for  a  single person to filter through all of the information and extract what they need. As a   result,   information   filtering   algorithms   are   necessary   to   uncover meaningful  information  from  the  massive  amount  of  data  already  available online.  Users  can  benefit  from  recommendation  systems  (RSs)  since  they simplify  the  process  of  identifying  relevant  information.  User  ratings  are incredibly  significant  for  creating  recommendations.  Previously,  academics relied   on   historical   user   ratings   to   predict   future   ratings,   but   today, consumers are paying more attention to user reviews because they contain so much relevant information about the user's decision. The proposed approach uses written testimonials to overcome the issue of doubt in the ratings' pasts. Using two data sets, we performed experimental evaluations of the proposed framework.  For  prediction,  clustering  algorithms  are  used  with  natural language processing in this strategy. It also evaluates the findings of various methods,   such   as   the   K-mean,   spectral,   and   hierarchical   clustering algorithms,  and  offers  conclusions  on  which  strategy  is  optimal  for  the supplied use cases. In addition, we demonstrate that the proposed technique outperforms alternatives that do not involve clustering.