Skip navigation
st. Mary's University Institutional Repository St. Mary's University Institutional Repository

Please use this identifier to cite or link to this item:
Authors: Mekonnen, Kalkidan
Keywords: opinions, Opinion Mining, Review, Sentence Level, Document Level, Feature Level, Classification, Extraction, Machine learning algorithms, Determination
Issue Date: Feb-2022
Abstract: Today, digital reviews play a pivotal role in enhancing global communications among consumers and influencing consumer buying patterns. The availability of technology and infrastructure create opportunities for citizens to publicly voice their opinions over social media. Business Company uses this opportunity to improve the quality of their product and the efficiency of their company. Companies don‟t yet have an effective way to make sense of customer opinions given on the product. Now a day‟s huge amount of product reviews are posted on the Web. Such a product reviews are a very important source of information for business companies to know about their product acceptance by their customer. Manual analysis of these reviews is very difficult because of the increase in the numbers of reviews on products day after day. Techno Company creates a Facebook page which helps consumers to share their experience and provide real insights about the performance of the product to future buyers. In order to extract valuable insights from a large set of reviews, classification of reviews and rating products into 1for best product which is highly accepted by their customer, 2 for good product and 3 for products having problem which customers is not happy to buy Product review Analysis is a computational study to extract subjective information from the text. This paper proposes a customer opinion analysis model to classify product reviews and rating the product best, good and bad based on the customer feedback about the product. It applies six popular machine learning classifiers namely: Support Vector Machine (SVM), BOOSTING, SLDA, NNETWOR, TREE and BAGGING with the aim to come up with the most efficient classifier. The dataset used consists of 2000 reviews about mobile phone products, collected from Tecno Facebook page. In order to evaluate the six classifiers, we used 10fold cross validation, recall, precision, F1-mesaure and accuracy to measure the performance of each algorithm. The results showed that SVM and BOOSTING outperformed the other classifiers in term of accuracy in all experiments. Decision Tree algorithm gave the lowest results across all experiments in terms of accuracy.
URI: .
Appears in Collections:Master of computer science

Files in This Item:
File Description SizeFormat 
Kalkidan Mekonnen Thesis.pdf1.2 MBAdobe PDFView/Open
Show full item record

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.