Thesis (MSc) - Cyprus International University. Institute of Graduate Studies and Research Computer Engineering Department

Includes bibliography (sheets 43-47)

ABSTRACT
Customer reviews of e-commerce service and products offering is a very critical and
important in the business world. The e-commerce industry is driven by enormous
data generation which is often collected for analysis and evaluation to gain both
insight into decision making for customer purchase and also by the business
organizations to enable them drive their businesses with informed decisions from
customer review data. machine learning and natural language processing has gained
so much technological development that it has been used in recent years to
significantly contribute to the research and development of sentiment analysis of
customer reviews and other forms of natural language processing. RNNs and LSTM
in particular have been identified through research literature as a suitable deep
machine learning models suitable for natural language processing due to their
recurrent and sequential processing of data because natural language is also a
sequential process. This study was carried out using LSTM recurrent neural network
for Amazon polarity dataset classification, the study also applied the use natural
language processing techniques of stemming and lemmatization which has shown
great performance improvement compared on the experiment results, the results of
this study was reported with an accuracy of 96%.
Keywords: Amazon Customer Reviews, Deep Learning, LSTM, Natural Language
Processing, Recurrent Neural Network