Engineering, Technology & Applied Science Research Vol. 13, No.

3, 2023, 10849-10855 10849

Sentiment Classification based on Machine

Learning Approaches in Amazon Product
Reviews
Mohammad Abu Kausar
Department of Information Systems, University of Nizwa, Oman
[email protected] (corresponding author)

Sallam Osman Fageeri

Department of Information Systems, University of Nizwa, Oman
[email protected]

Arockiasamy Soosaimanickam
Department of Information Systems, University of Nizwa, Oman
[email protected]
Received: 15 March 2023 | Revised: 14 April 2023 | Accepted: 23 April 2023
Licensed under a CC-BY 4.0 license | Copyright (c) by the authors | DOI: https://doi.org/10.48084/etasr.5854

ABSTRACT
Online retailers and merchants increasingly request feedback from their clients on the products they
purchase. This has led to a significant increase in the number of product reviews posted online, as more
people are making purchases online. The opinions expressed in these customer reviews have a significant
impact on other customers' purchase decisions, as they are influenced by other customers'
recommendations or complaints. This study used Amazon, a well-known and widely used e-commerce
platform, to examine sentiment categorization using several machine learning techniques while analyzing
an Amazon Reviews dataset. At first, the reviews were transformed into vector representations using the
Bag-of-Words approach. Word cloud was used to illustrate the text data in terms of the frequency they
appear in the review. Subsequently, the machine learning methods decision trees and logistic regression
were used. The two models used in this study achieved high levels of accuracy in analyzing the dataset.
Specifically, the Decision Tree model outperformed the Logistic Regression one, achieving an impressive
accuracy of 99% compared to the 94% of the latter.

Keywords-sentiment analysis; Amazon customer reviews; dataset; feature extraction; text classification;
machine learning

I. INTRODUCTION Sentiment analysis and classification is a field of computer

science that attempts to answer this problem by sifting through
Over the past decades online marketplaces have grown in given natural language texts to extract attitudes and views,
popularity, leading to the trend of asking customers for using numerous approaches, such as biometrics, computational
feedback on purchased products to improve the overall linguistics, natural language processing, and text analysis. Due
customer experience. Every day, millions of reviews about to their accuracy and effectiveness, Machine Learning (ML)
different goods, services, and locations are created online. As a approaches have lately grown in favor of the semantic and
result, the Internet has become the key resource for knowledge review analysis fields. This study considered Amazon as one of
and opinions regarding products and services. However, the the most popular e-commerce vendors. Potential customers
vast amount of product reviews and the multiple perspectives may read hundreds of reviews posted by previous customers
to consider may further complicate the decision-making regarding the products they're interested in [1-2]. These
process, leading to confusion and uncertainty, as it is difficult reviews provide valuable information on a product, helping
for customers to choose wisely when there are several shoppers understand most of its features. Customers benefit
viewpoints on the same product and different ratings. from this, but it also helps retailers or producers to get a greater
Therefore, e-commerce businesses should evaluate this content knowledge of customers' needs.
to get feedback on their products and help customers decide.
This study used supervised algorithms to address the
problem of sentiment classification for online reviews and

www.etasr.com Kausar et al.: Sentiment Classification based on Machine Learning Approaches in Amazon Product …
 Engineering, Technology & Applied Science Research Vol. 13, No. 3, 2023, 10849-10855 10850

determine the overall significance of customer evaluations and analysis to improve company operations and client retention, as
characterize them as positive or negative. The data used were analyzing product evaluations allows a company to understand
reviews of Amazon Titan Men Watches, collected from client experiences [16]. A client can leave a review to indicate
amazon.in. The study employed supervised algorithms to whether he is happy or unhappy with a particular product or
address the issue of sentiment classification for online reviews. service. However, the majority of product reviews do not
The goal was to determine the overall importance of customer represent the level of client happiness. In [16], a study was
evaluations by categorizing them as either positive or negative. conducted based on reviews on the Internet to classify
consumer happiness, according to auditory and language
II. RELATED WORKS characteristics, into four classes: extremely positive, positive,
Many studies used data collected from numerous sources, neutral, or highly unfavorable. The findings of this study on the
such as Twitter [3-4], product reviews, consumer feedback, etc. impact of review length on online sentiments were consistent
Customer engagement programs help businesses to strengthen with [17]. Customers prefer to rely on comprehensive and
their emotional relationships with clients from all over the insightful product reviews before making a final purchase [18].
world. Customer involvement is also influenced by product As a result, companies must discover the true degree of
reviews [5]. The study of product reviews allows a company to customer satisfaction with their products and services in order
learn how customers feel about its products. According to [6], to make an informed decision based on online product reviews.
online consumer reviews have an important influence in Authors in [19] evaluated the way internet reviews
molding customers' online shopping decisions. The term influence Amazon book sales. According to this study,
"review" refers to the process of determining whether a product customers view online reviews as a reliable source of
or service is suitable for a certain purpose. Every day, a information and prefer more accessible and detailed reviews.
massive quantity of fresh information is added to the Web [7], According to this study, internet reviews have a major impact
hence a method was proposed based on simultaneous web on user experiences and product costs. These findings were in
crawling using mobile agents. agreement with those of [20] on online reviews and feelings.
Sentiment analysis is used by businesses to increase their The study also looked at the valence of internet reviews. In
competitiveness in the market, as it enables them to [19], the influence of online reviews on purchases was found to
comprehend the opinions and experiences of their clients on be contradictory. Some studies concluded that the valance of
their products and services [8]. Sentiment analysis is also used online reviews has a major influence on sales, while others
to capture market sentiments to design a better forecasting found that it has no impact. The impact is also affected by
model for the stock market [9]. Modern organizations use variables such as product categories and qualitative text
sentiment analysis to improve their word-of-mouth marketing qualities. In [21], an analysis was conducted on 142.8 million
approach. Competitive companies use text-mining tools to Amazon user reviews, focusing on determining the usefulness
better understand client experiences and extract valuable and unhelpfulness of each review by examining the summary
information from social networking platforms, newspapers, and headline, the product remark, and helpfulness information,
other sources. In [10], k-Nearest Neighbors (k-NN), Decision filtering out blank and non-English product evaluations to
Trees (DT), and Artificial Neural Network (ANN) models were improve the accuracy of the results, and choosing only those
used to identify client behavior in the banking industry. As a with the most votes. The study concluded that an investigation
result, a mix of sentiment analysis, text mining, and other of online product evaluations on Amazon plays an important
methods is critical in ensuring that businesses understand and part in today's e-commerce. Helpful reviews provide thorough
capitalize on online consumer evaluations. Sentiment information on specific products or services based on client
extraction is an excellent method of understanding customer feedback [22]. Customers rely on reviews with the most votes
assumptions online [11], as the data collected from internet to make a purchase decision. The results are congruent with the
platforms and product review sites allow businesses to improve findings of [23] on how customer reviews impact internet
their marketing methods. Product reviews also influence client purchases. Positive product reviews enable clients to build
purchase decisions. Competitive organizations, such as greater trust in the things they want to buy online.
Amazon, use such information to make decisions [12], and Sentiment analysis has gaps due to issues with the accuracy
leading merchants in the United States rely on Internet product and reliability of the utilized methods, as well as the lack of
reviews to improve their marketing efforts and company standardization in sentiment interpretation and classification.
procedures [13]. For example, if a given product receives a More research is needed to optimize customer engagement
large number of unfavorable reviews for its cost or quality, the programs and improve the understanding of customer
company examines the problem and tries to address it as soon sentiments. This study aims to address the potential gaps in the
as possible. sentiment analysis based on Machine Learning Approaches in
According to [6], most of the product evaluations on the Amazon Product Reviews using Decision Trees and Regression
Internet are less polar and more positively balanced. Also, the models. Specifically, this study investigated the effectiveness
distribution of product reviews for similar items differs by of different customer engagement strategies in influencing
platform [14]. The diversity of reviews is driven by several customer sentiments and explored the impact of these strategies
factors, including the rating system, the business strategy of the on sentiment classification accuracy.
online platform, and the review frequency [6]. In [15], a Web
Crawling Model based on Java Aglets was presented.
Furthermore, it is important to mention that firms use sentiment

www.etasr.com Kausar et al.: Sentiment Classification based on Machine Learning Approaches in Amazon Product …
 Engineering, Technology & Applied Science Research Vol. 13, No. 3, 2023, 10849-10855 10851

III. METHODOLOGY
Sentiment analysis is considered a study of people's
sentiments, feelings, and views as conveyed via writing, and is
useful in understanding other peoples' points of view on any
subject. Opinions can be positive, negative, or neutral. The
suggested method was based on an ML prediction model, to
analyze both positive and negative reviews by binary
classification. Figure 1 shows the steps of this method,
beginning with data collection and ending with the evaluation
of each classification model.
Fig. 2. Actual Amazon customer review sample.

Data collection was the initial stage of the study. Raw data
were obtained from the website amazon.in and converted to
Comma Separated Values (CSV) format. The data set
contained 4960 Titan Men Watches reviews.
C. Data Cleaning
This is a critical stage in examining any type of data and
has a significant influence on the success of ML models. There
are numerous types of pre-processing procedures and the
appropriate ways must be selected. This study employed four
distinct data preparation steps in the reviews:
1. Remove Emojis.
2. Remove HTML tags.
3. Lowercase all letters.
4. Filter numbers and special characters.
1) Emoji Removal
Although people use emojis to express their feelings, they
were removed since they did not affect the identification of the
polarity of the review.
2) HTML Tag Removal
Fig. 1. Overall sentiment analysis approach for Amazon reviews. The HTML tags of the retrieved reviews were removed
because they did not affect the determination of the polarity of
A. Programming Environment the review.
Python is one of the most used programming languages in 3) Converting all Letters to Lowercase
data science and ML, as it offers a large library collection for
In different reviews, there is a good chance that identical
solving various ML problems. Python was chosen due to its
words will appear in different situations and the system will
extensive libraries and ease of use. Scikit-learn is a Python
recognize them as distinct words. Converting all letters to
package that provides supervised ML algorithms [24], lowercase was used to avoid such problems.
including many classification algorithms, such as SVM and
Naive Bayes, and feature extraction techniques. 4) Filtering Numbers and Special Characters
B. The Dataset All the unnecessary elements in determining the review's
polarity were eliminated to make the data tidy and clean. The
Figure 2 presents an example of an Amazon review to
special letters and digits were eliminated.
better grasp the dataset's structure and format. An Amazon user
review consists of the following four key components that D. Feature Extraction
assist in comprehending and analyzing the reviews: ML algorithms interpret data in specified formats. The text
 Summary: The title of the review. data were turned into numerical feature vectors, a process
known as vectorization. Bag of Words is one such approach
 Review text: The review's actual content. that involves tokenization, normalization, and counting. This
 Rating: The product's user rating on a scale of 1 to 5. study used CountVectorizer to represent words in terms of Bag
of Words. CountVectorizer requires specifying an N-gram
 Helpfulness: The percentage of persons who considered the range, which is a tuple consisting of the lowest and maximum
review beneficial. length of the sequence of words to be regarded as features.

www.etasr.com Kausar et al.: Sentiment Classification based on Machine Learning Approaches in Amazon Product …
 Engineering, Technology & Applied Science Research Vol. 13, No. 3, 2023, 10849-10855 10852

E. Word Cloud of Reviews

The word cloud is a method of displaying text data where
each word's size corresponds to its frequency or importance. A
word cloud can be used to highlight significant textual
information, and they are widely used in social network data
analysis. This study used word clouds to represent the most
often-used terms in reviews.
IV. EXPLORATORY DATA ANALYSIS
Exploratory Data Analysis (EDA) is a method of displaying
and evaluating data concealed in rows and columns, using
several charts to provide maximum insight into the dataset.
This study investigated the Positive, Negative, and Neutral
review categories of Amazon reviews based on the sentiment
score. Figure 3 shows the Amazon product review categories.
This study scrapped 4960 reviews of a specific product and
stored them, after cleaning, in a CSV file.

Fig. 5. Most common words used in positive reviews

Fig. 3. Distribution of reviews' categories.

The text data were presented graphically using Python's

Word Cloud package. Figures 4-6 show the most common
words used in reviews, as positive reviews, and as negative
reviews, respectively.

Fig. 6. Most common words in negative reviews.

A. Unigrams
Unigrams determine the frequency of single words in the
reviews. Figures 7-8 show the unigrams of the top 20 positive
and negative reviews, respectively.
B. Bigrams
Unigrams do not offer a clear understanding of a
consumer's intended message. Bigrams can capture the
meaning behind consecutive pairs of words in the reviews.
Combining two adjacent words into a single unit can offer
valuable insight into the context and meaning of a text.
Bigrams show the frequency of two-word combinations in a
text review. For instance, in the sentence "I love this product",
the bigram "love this" would be generated, while "I hate this
product" would produce the bigram "hate this". Analyzing the
most prevalent bigrams in a collection of reviews can help to
Fig. 4. Most common words used in the reviews. understand what consumers are attempting to convey regarding
their encounter with a product or service.

www.etasr.com Kausar et al.: Sentiment Classification based on Machine Learning Approaches in Amazon Product …
Engineering, Technology & Applied Science Research Vol. 13, No. 3, 2023, 10849-10855 10853

Fig. 7. Top 20 positive review unigram. Fig. 8. Top 20 negative review unigram.

Fig. 9. Top 40 positive and negative review bigrams.

www.etasr.com Kausar et al.: Sentiment Classification based on Machine Learning Approaches in Amazon Product …
 Engineering, Technology & Applied Science Research Vol. 13, No. 3, 2023, 10849-10855 10854

Figure 9 shows the top 40 positive and negative review

bigrams. The bar plot shows that the most common bigrams in
positive reviews were "value for money", "lovely watch",
"good product", etc. These bigrams suggest that customers
were satisfied with the quality and value of the product they
purchased. In contrast, the most commonly used bigrams in
negative reviews were "media could", "could not", "titan
watch", etc., implying that consumers experienced problems
with the product, such as difficulty with media playback or
loading times. Overall, bigrams are a useful tool in natural
language processing and sentiment analysis, providing valuable
insight into the opinions and experiences of consumers. By
Fig. 10. Accuracy comparison for the methods used.
identifying and analyzing the most common bigrams in a set of
reviews, a better understanding of the strengths and weaknesses
B. Confusion Matrix
of a product or service can be gained. This information can
then be used to make more informed decisions. It is worth noting that the Accuracy values reported are not
the only performance metrics to consider when evaluating the
V. RESULTS AND ANALYSIS models. Other metrics such as Precision, Recall, and F1-score
This study used Python and Jupyter, in conjunction with can provide a more nuanced view of the models' performance
supporting libraries, to perform data purification, visualization, and should be considered alongside Accuracy. They can be
pre-processing, and ML modeling. Supervised ML was used to used to assess the performance of the categorization methods.
create sentiment classification models, by building training and These parameters are important for measuring the efficacy of
testing sets. A collection of features was then taken from the supervised ML algorithms since they are based on the
training and testing data and supplied into a classifier model, confusion matrix [25]. A confusion matrix is widely used to
such as Logistic Regression (LR) and Decision Tree (DT). The visualize an algorithm's performance. Classification terms TP,
complete review dataset was separated into two parts: 75% of TN, FN, and FP are used to compare class labels in this matrix,
the data was used to train the models, while 25% was used to as shown in Table II and Figure 11. Based on the data of the
assess their performance. The performance of the models was confusion matrix, Precision, Recall, F-measure, and Accuracy
validated using accuracy and confusion matrix. can be used to evaluate the performance of the classifier.

A. Accuracy (A) TABLE II. CONFUSION MATRIX

Accuracy (A) is defined as the proportion of correctly PV
Predicted-Negative Predicted-Positive
predicted occurrences to the total number of occurrences by: AV
Actual Negative TN FP
TP TN Actual Positive FN TP
A
TP TN FP FN
where TP stands for True Positive, TN for True Negative, FP
for False Positive, and FN for False Negative. Table I shows
the accuracy results of the models used in this study.

TABLE I. ACCURACY OF MODELS

Model Accuracy
Logistic Regression 94%
Decision Tree 99%

The results show that both Logistic Regression and

Decision Tree were effective. However, the two models had
different levels of accuracy, as DT achieved 99%
outperforming LR, which achieved 94%, as shown in Table I
and Figure 10. This could be attributed to DT's ability to handle
complex decision-making processes that may not be captured
by LR. The difference in accuracy between the two models is
significant, and it may impact the model selection decision
depending on the importance of the classification task. For
example, in a high-stakes classification task, such as in medical
diagnosis, the higher accuracy of DTs could be crucial to make
accurate and reliable predictions.
Fig. 11. The confusion Matrix of the LR model.

www.etasr.com Kausar et al.: Sentiment Classification based on Machine Learning Approaches in Amazon Product …
 Engineering, Technology & Applied Science Research Vol. 13, No. 3, 2023, 10849-10855 10855

VI. CONCLUSION [12] J. Lim, M. Park, S. Anitsal, M. M. Anitsal, and I. Anitsal, "Retail
Customer Sentiment Analysis: Customers’ Reviews of Top Ten U.S.
Sentiment analysis is a necessary and popular technique for Retailers’ Performance," Global Journal of Managment and Marketing,
collecting information from text data on e-commerce websites. vol. 3, no. 1, pp. 124–150, 2019.
E-commerce platforms produce enormous volumes of text data [13] R. S. Jagdale, V. S. Shirsat, and S. N. Deshmukh, "Sentiment Analysis
every day in the form of suggestions, reviews, tweets, and on Product Reviews Using Machine Learning Techniques," in Cognitive
Informatics and Soft Computing, Singapore, 2019, pp. 639–647,
comments. Additionally, emoticons, ratings, and reviews all https://doi.org/10.1007/978-981-13-0617-4_61.
suggest people's opinions. A customer can learn more about a
[14] V. Vyas and V. Uma, "Approaches to Sentiment Analysis on Product
product and make an informed choice by extrapolating Reviews," in Sentiment Analysis and Knowledge Discovery in
information from reviews. This study used multiclass and Contemporary Business, IGI Global, 2019, pp. 15–30.
binary classification for Amazon reviews of a product using [15] Md. A. Kausar, V. S. Dhaka, and S. K. Singh, "Web Crawler Based on
supervised machine learning methods such as Logistic Mobile Agent and Java Aglets," International Journal of Information
Regression and Decision Tree. Logistic Regression produced Technology and Computer Science, vol. 5, no. 10, pp. 85–91, Sep. 2013,
an excellent outcome with 94% accuracy and Decision Tree https://doi.org/10.5815/ijitcs.2013.10.09.
produced outstanding results with 99% accuracy. E-commerce [16] S. Govindaraj and K. Gopalakrishnan, "Intensified Sentiment Analysis
of Customer Product Reviews Using Acoustic and Textual Features,"
websites should consider various feature extraction methods ETRI Journal, vol. 38, no. 3, pp. 494–501, 2016,
and machine learning techniques, examine additional product https://doi.org/10.4218/etrij.16.0115.0684.
categories, analyze unstructured data, and incorporate [17] M. Ghasemaghaei, S. P. Eslami, K. Deal, and K. Hassanein, "Reviews’
sentiment analysis into customer experience strategies to length and sentiment as correlates of online reviews’ ratings," Internet
enhance customer satisfaction and loyalty. Research, vol. 28, no. 3, pp. 544–563, Jan. 2018,
https://doi.org/10.1108/IntR-12-2016-0394.
REFERENCES [18] P. Sasikala and L. Mary Immaculate Sheela, "Sentiment analysis of
online product reviews using DLMNN and future prediction of online
[1] X. Fang and J. Zhan, "Sentiment analysis using product review data,"
product using IANFIS," Journal of Big Data, vol. 7, no. 1, May 2020,
Journal of Big Data, vol. 2, no. 1, Jun. 2015, Art. no. 5, https://doi.org/
Art. no. 33, https://doi.org/10.1186/s40537-020-00308-7.
10.1186/s40537-015-0015-2.
[19] S. K. Sharma, S. Chakraborti, and T. Jha, "Analysis of book sales
[2] J. McAuley, "Amazon product data," Recommender Systems and
prediction at Amazon marketplace in India: a machine learning
Personalization Datasets. https://cseweb.ucsd.edu/~jmcauley/datasets.
approach," Information Systems and e-Business Management, vol. 17,
html#amazon_reviews.
no. 2, pp. 261–284, Dec. 2019, https://doi.org/10.1007/s10257-019-
[3] M. A. Kausar, A. Soosaimanicka, and M. Nasar, "Public Sentiment 00438-3.
Analysis on Twitter Data during COVID-19 Outbreak," International
[20] A. Y. L. Chong, B. Li, E. W. T. Ngai, E. Ch’ng, and F. Lee, "Predicting
Journal of Advanced Computer Science and Applications, vol. 12, no. 2,
online product sales via online reviews, sentiments, and promotion
2021, https://doi.org/10.14569/IJACSA.2021.0120252.
strategies: A big data architecture and neural network approach,"
[4] M. Mahyoob, J. Algaraady, M. Alrahiali, and A. Alblwi, "Sentiment International Journal of Operations & Production Management, vol. 36,
Analysis of Public Tweets Towards the Emergence of SARS-CoV-2 no. 4, pp. 358–383, Jan. 2016, https://doi.org/10.1108/IJOPM-03-2015-
Omicron Variant: A Social Media Analytics Framework," Engineering, 0151.
Technology & Applied Science Research, vol. 12, no. 3, pp. 8525–8531,
[21] J. Du, J. Rong, S. Michalska, H. Wang, and Y. Zhang, "Feature selection
Jun. 2022, https://doi.org/10.48084/etasr.4865.
for helpfulness prediction of online product reviews: An empirical
[5] N. Nandal, R. Tanwar, and J. Pruthi, "Machine learning based aspect study," PLOS ONE, vol. 14, no. 12, 2019, Art. no. e0226902,
level sentiment analysis for Amazon products," Spatial Information https://doi.org/10.1371/journal.pone.0226902.
Research, vol. 28, no. 5, pp. 601–607, Oct. 2020, https://doi.org/
[22] Meenakshi, A. Banerjee, N. Intwala, and V. Sawant, "Sentiment
10.1007/s41324-020-00320-2.
Analysis of Amazon Mobile Reviews," in ICT Systems and
[6] V. Schoenmueller, O. Netzer, and F. Stahl, "The Polarity of Online Sustainability, Singapore, 2020, pp. 43–52, https://doi.org/10.1007/978-
Reviews: Prevalence, Drivers and Implications," Journal of Marketing 981-15-0936-0_4.
Research, vol. 57, no. 5, pp. 853–877, Oct. 2020, https://doi.org/
[23] K. Q. Anh, Y. Nagai, and L. M. Nguyen, "Extracting Customer Reviews
10.1177/0022243720941832.
from Online Shopping and Its Perspective on Product Design," Vietnam
[7] Md. A. Kausar, V. S. Dhaka, and S. K. Singh, "An Effective Parallel Journal of Computer Science, vol. 06, no. 01, pp. 43–56, Feb. 2019,
Web Crawler based on Mobile Agent and Incremental Crawling," https://doi.org/10.1142/S2196888819500088.
Journal of Industrial and Intelligent Information, vol. 1, no. 2, pp. 86–
[24] F. Pedregosa et al., "Scikit-learn: Machine Learning in Python," The
90, Jun. 2013, https://doi.org/10.12720/jiii.1.2.86-90.
Journal of Machine Learning Research, vol. 12, pp. 2825–2830, 2011.
[8] I. Karamitsos, S. Albarhami, and C. Apostolopoulos, "Tweet Sentiment
[25] A. Tripathy, A. Agrawal, and S. K. Rath, "Classification of sentiment
Analysis (TSA) for Cloud Providers Using Classification Algorithms
reviews using n-gram machine learning approach," Expert Systems with
and Latent Semantic Analysis," Journal of Data Analysis and
Applications, vol. 57, pp. 117–126, Sep. 2016, https://doi.org/10.1016/
Information Processing, vol. 7, no. 4, Nov. 2019, Art. no. 69212,
j.eswa.2016.03.028.
https://doi.org/10.4236/jdaip.2019.74016.
[9] U. P. Gurav and S. Kotrappa, "Sentiment Aware Stock Price Forecasting
using an SA-RNN-LBL Learning Model," Engineering, Technology &
Applied Science Research, vol. 10, no. 5, pp. 6356–6361, Oct. 2020,
https://doi.org/10.48084/etasr.3805.
[10] A. Rahman and M. N. A. Khan, "A Classification Based Model to
Assess Customer Behavior in Banking Sector," Engineering, Technology
& Applied Science Research, vol. 8, no. 3, pp. 2949–2953, Jun. 2018,
https://doi.org/10.48084/etasr.1917.
[11] V. K. Jain, S. Kumar, and P. Mahanti, "Sentiment Recognition in
Customer Reviews Using Deep Learning," International Journal of
Enterprise Information Systems (IJEIS), vol. 14, no. 2, pp. 77–86, Apr.
2018, https://doi.org/10.4018/IJEIS.2018040105.

www.etasr.com Kausar et al.: Sentiment Classification based on Machine Learning Approaches in Amazon Product …