Sentiment Analysis of Consumer Reviews: Unveiling Perspectives and Building a Machine Learning Model for Product Evaluation


  • Hassan Latif Department of Computer Science, Bahauddin Zakariya University, Multan
  • Muhammad Imran Department of Computer Science, Bahauddin Zakariya University, Multan
  • Rabia Javed Department of Computer Science, TIMES Institute, Multan, Pakistan.
  • Adil Siddique Department of Computer Science, NUML, Multan, Pakistan.



Sentiment Analysis, Machine Learning, Natural Language Processing


The goal of this paper is to classify the unusual and dreadful critiques of the clients over one-of-a-kind products and put up a control analysis version to polarize big amounts of analysis. Our dataset comprises consumer reviews and ratings, which are information we have obtained through user evaluations of Amazon goods. We took the skills from our dataset and developed a large supervised version entirely based on them. These modes not great consist of conventional algorithms mutually with naive bays, linear helping vector machines, and k-nearest fellow resident, but also deep studying metrics which encompass frequent Neural Networks and complication neural networks. We evaluated the correctness of those fashions and got a higher know-how of the inverse attitude earlier to the produce. The initial objective is to describe the situation from the client's perspective and assess the intensity of the emotion. The goal of the second assignment is to build and train a machine learning system that can be used to divide customer evaluations into two categories: excellent and terrible. Regardless of the information that Amazon does now not have an API like Twitter to download evaluation with, it does have links for every assessment on every item for consumption, to theoretically traverse the web page through product IDs. We used Perl script written by way of Andrea Eula to get hold of the evaluation for the stimulation and some poles apart products. The most important script downloads the entire HTML page for the product, and the second searches the record for data about the appraisal, such as the product ID, rating, assessment date, and estimate passage. This study shows that the model had a classification accuracy of 78% and a precision similar to the sympathy and genuine score, but the Kappa do curve was rather low. In the future, more work can be done by improving the key parameters.