=Paper=
{{Paper
|id=Vol-2853/paper31
|storemode=property
|title=Intelligent Method of a Competitive Product Choosing based on the Emotional Feedbacks Coloring
|pdfUrl=https://ceur-ws.org/Vol-2853/paper31.pdf
|volume=Vol-2853
|authors=Roman Gramyak,Hrystyna Lipyanina-Goncharenko,Anatoliy Sachenko,Taras Lendyuk,Diana Zahorodnia
|dblpUrl=https://dblp.org/rec/conf/intelitsis/GramyakLSLZ21
}}
==Intelligent Method of a Competitive Product Choosing based on the Emotional Feedbacks Coloring==
https://ceur-ws.org/Vol-2853/paper31.pdf
Intelligent Method of a Competitive Product Choosing based on
the Emotional Feedbacks Coloring
Roman Gramyaka, Hrystyna Lipyanina-Goncharenkoa, Anatoliy Sachenkoa,b, Taras Lendyuka
and Diana Zahorodniaa
a
West Ukrainian National University, Lvivska str., 11, Ternopil, 46000, Ukraine
b
Kazimierz Pulaski University of Technology and Humanities in Radom, Department of Informatics, Jacek
Malczewski str., 29, Radom, 26 600, Poland
Abstract
Finding the best products for sale is one of the most important steps in the process of a
profitable company creating. That is why the choice of goods for the online store must be
made carefully, taking into account both the opportunities and analysis of prospects in the
niche, and a number of other important parameters. One of the methods of competitive
product choosing can be products analysis in marketplaces based on the emotional feedbacks
coloring. Research on product feedbacks is an extremely popular topic, as confirmed by
research analysis. Feedbacks can be constantly re-read, but when there are many products in
one segment, because there are more and more manufacturers, it is time consuming.
Therefore, the development of technology that can automate this process is necessary for the
sales business. Paper develops the intelligent method of a competitive product choosing
based on the emotional feedbacks coloring, which is divided into three blocks: parser of
feedbacks, emotional coloring determination and feedbacks classification. The data will help
retailers manage their websites wisely and help customers make purchasing decisions. The
implementation of the method was carried out on the data of the Ukrainian site Rozetka,
where 4477 feedbacks were used. The classification was tested by eight classical machine-
based classification methods, namely Support Vector Classifier, Stochastic Gradient Decent
Classifier, Random Forest Classifier, Decision Tree Classifier, Gaussian Naive Bayes, K-
Neighbors Classifier, Ada Boost Classifier, Logistic Regression.
Keywords 1
Product, feedback, parser, text emotional coloring, classification, machine learning.
1. Introduction
Other people’s opinions have always been an important piece of information for most of us in the
decision-making process. The interest shown by users in online feedback and comments, as well as
the potential impact of these comments on issues in discourse and decision-making, make us pay
attention to this aspect of online activity. The purpose of tone analysis is to find ideas in the text and
determine their properties. This method finds its practical application in such fields as sociology,
political science, marketing, medicine and many others.
The main approaches that can be used to analyze emotional mood are machine learning and a
vocabulary-based approach. Currently, the analysis of attitudes is an interesting topic and direction of
development, as it has many practical applications. Companies use it to automatically analyze survey
feedbacks, product feedbacks, and social media comments to gain valuable information about their
brands, products, and services. However, today seller can use the analysis of emotional mood and to
IntelITSIS’2021: 2nd International Workshop on Intelligent Information Technologies and Systems of Information Security, March 24–26,
2021, Khmelnytskyi, Ukraine
EMAIL: fear3171@gmail.com (R. Gramyak); xrustya.com@gmail.com (H. Lipyanina-Goncharenko); as@wunu.edu.ua (A. Sachenko);
tl@wunu.edu.ua (T. Lendyuk); dza@wunu.edu.ua (D. Zahorodnia)
ORCID: 0000-0001-8698-0377 (R. Gramyak); 0000-0002-2441-6292 (H. Lipyanina-Goncharenko); 0000-0002-0907-3682 (A. Sachenko);
0000-0001-9484-8333 (T. Lendyuk); 0000-0002-9764-3672 (D. Zahorodnia)
© 2021 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
select the potentially most profitable product that is in good demand in the market, it will allow online
sellers to reduce the risks of choosing a product for trade, and thus increase revenue.
In this regard, it can be considered that the development of a method of selection by machine
learning algorithms of the intelligent method of a competitive product choosing based on the
emotional feedbacks coloring is one of the most promising areas in online commerce.
This paper is structured as follows. A Section 2 discusses the analysis of related work, and the
Section 3 presents the proposed intelligent method of a competitive product choosing based on the
emotional feedbacks coloring. The Section 4 presents a case study and discussion, and the Section 5 is
summarizing the obtained results.
2. Related works
Reviewing customer feedbacks in online stores is attracting more and more attention from
practitioners and scientists. Many studies have analyzed the impact of online feedbacks on sales of
goods and services: Amazon products [1], technical characteristics of products from social networks
[8], Flipkart, Snapdeal and Amazon India brands [34] and Amazon, Sony and Pocketbook on
Facebook, reviews of restaurants from Yelp.com [49], reviews of books (book.dangdang.com). Also,
there are works that analyze research on predicting the usefulness of the review [40], identifying data
sources, ML methods [23].
Important after data retrieval is the ranking of these data, [18] presents the structure of merging
information to rank product feedbacks. Also, ranking approaches are proposed: method on different
aspects of alternative products [11]; model [19] of regression prediction based on their quality;
hierarchical approach [20], word level and review level; classifier [15] Ensemble, which uses linguistic
features to praise or classify complaints; method [21], which returns the ranking lists of useful
feedbacks according to their usefulness in relation to the product, a model of multiple linear regression
using the method of elastic network regularization.
After structured feedback receiving, seller need to conduct a detailed analysis. In [3] there is
developed the system for predicting the usefulness of the feedback for multilingual online feedbacks,
which displays the best results in terms of forecasting the usefulness and classification. A [13] model
for detecting product defects based on feedback on social networks is proposed. In [14] the method of
analysis of competitive advantages of the product is offered, which provides an important basis for
quality management and development of marketing strategy by UGC mining. An empirical study of
the choice of functions for predicting the usefulness of online product feedbacks was conducted in
[22]. The relationship between moods and usefulness for feedback was studied [26], a method was
developed for full use of mood features in assessing usefulness for feedback and it was investigated
whether the type of product affects the assessment of usefulness for feedback. Based on the bias of
negativity and the theory of summation, [33] proposed a theoretical model that explains the usefulness
of feedbacks on the Internet based on the specific characteristics of these feedbacks (i.e., duration,
evaluation, arguments frame). [43] developed a system that takes all feedbacks containing Hindi, as
well as texts in English, and determines the mood expressed in this feedback for each attribute of the
product, as well as the final feedback of the product. Several utility predictions models have been
studied [48] using multidimensional adaptive regression, classification and regression trees, random
forest approaches, neural networks, and deep neural networks using two real Amazon product feedback
datasets.
Determining the emotional state is the next step in analyzing user feedback. To analyze the
emotional mood of feedbacks, seller can use several approaches, namely: models [30, 32] of multiple
regression; ensemble method [9] with several revisions; clustering of text [6] focused on the business
sector; model [4] consumer decision-making in terms of risk; method [16] to dynamically assign a
coefficient of influence to each basic student in the ensemble; SentiCon by LSA consolidation [25];
model [29] prediction based on a combination of neural network (CNN) and TransE; method [38] of
classification of action verbs in the feedback text; integrated model [35] of decision support for finding
products on the Internet; approach [44] is based on language modeling (LM); model [7] of product
recommendations based on product rating, filtering and hybrid classification of decision trees; method
[10] of hybrid classification of moods on the basis of attributes for the purpose of receiving orientation
on moods, approach [12] on the basis of fuzzy numbers of an interval of type 2; model [31] predicting
the usefulness of feedback is presented using a truncated decision tree C4.5.
The most accurate in determining the emotional mood of the feedback are models [5, 36, 37] based
on machine learning, deep learning [17, 39] and convolutional neural network [2]. An uncontrolled
thematic model of emotion extraction has been proposed [24], which adds appropriate relationships
between aspect words and thought words to express comments as a bag of aspect-thought pairs. A
model is presented [27] for determining the usefulness of comments using text functions removed from
feedbacks, based on different types of functions, including emotional, linguistic and textual features,
values of valence, excitation and dominance, feedback duration and polarity of comments. An in-depth
model has been proposed [28] for using the features of feedbacks based on content, semantics,
attitudes, and metadata to predict the usefulness of the feedback. [41] presents a general structure that
uses natural language processing (NLP) methods, including mood analysis, text data analysis, and
clustering methods, to obtain new estimates based on consumer sentiment for different product
characteristics. It is proposed [42] to use the plot of the film Wikipedia to identify genre factions using
methods of text extraction, based on the model of word reasoning, where (frequency) of occurrence of
each word is used as a feature to prepare a classifier. [45] various model architectures have been
proposed that can be used to predict the genre and rating of a film in the different languages available
in our abstract-based data set. Proposed [46] approach to the classification of moods and tweet
language, the architecture includes a convolutional neural network (ConvNet) with two different
outputs, each of which is designed to minimize the error of classification or assignment of distribution,
or language identification. [47] The classification of the film genre is proposed only on the basis of
poster images, on the basis of an in-depth neural network.
As it confirmed by the analysis above a research on product feedbacks is an extremely popular
topic, the data help retailers manage their websites wisely and help them make purchasing decisions.
Analysis of the emotional state of the feedbacks is an important point in solving the problem of
determining the most popular product, which allows the seller to choose the most profitable product
for sale.
In this regard, a goal of this paper is to develop the intelligent method of a competitive product
choosing based on the emotional feedbacks coloring.
3. Proposed method
To indicate further the novelty of the proposed method authors conducted a detailed analysis of
analogues where data is extracted, the emotional mood of the feedbacks is determined and classified
(Table 1).
Table 1
Analysis of research analogues
Formulation of the Data Emotional Classification
Source Data
problem mining coloring Method Accuracy
6 A model of product DOM Normalized Random 0.9691 www.amazon.com
recommendations based Parser Product Tree 0.978
on mood assessments; Feedback Score Hoeffding 0.967
real-time product data is Tree
proposed Adaboost
+RT
11 A new ranking method is - LDA PageRank - -
offered through online
feedbacks based on
various aspects of
alternative products,
which combines both
objective and subjective
values.
14 A method of analysis of Parser UGC NTUSD 0.7425 Social networks
competitive advantages Hownet 0.6425
Formulation of the Data Emotional Classification
Source Data
problem mining coloring Method Accuracy
of the product is Trained
proposed, which domain- 0.8625
provides an important specific
basis for quality sentiment
management and lexicon with
development of H=9
marketing strategy
through the extraction of
UGC.
18 The approach of Parser LSGDM SVM 0.63 www.twitter.com
networks to the
detection of social SVM-DS 0.97
networks based on the
method of fuzzy
clustering.
21 A method with an DOM LS2F Decision 0.77 www.amazon.com
extended list of tree with
functions for presenting LS2F
the overview and a
model of multiple linear
regression using the
method of elastic
network regularization is
proposed.
26 The data set used - Python Random 0.5742 www.amazon.com
methods to obtain library forest
information and assess
moods. Gradient SentiWordNet Gradient 0.5721
amplification and F- boosting
estimation methods
were used to classify the
data set with these
characteristics.
33 ANOVA assays are used - ANOVA ANN 0.84 -
to identify the levels of
each of these
characteristics that
maximize the tangible
usefulness of online
consumer feedbacks,
and an artificial neural
network approach is
used to predict the
usefulness of this
feedback based on its
characteristics.
37 A new approach called - - Classification 0.706 JD.com
value-based neighbor threshold
selection (VNS) is
Formulation of the Data Emotional Classification
Source Data
problem mining coloring Method Accuracy
proposed to address the
above constraints.
39 The OCC model and the - OCC CNN 0.84 Chinese social
CNN-based networks
generalization method
for Chinese
microblogging systems
are proposed.
41 A general structure that - -NLP RSS - www.amazon.com
uses natural language
processing techniques,
including mood analysis,
text data analysis, and
clustering techniques, to
obtain new estimates
based on consumer
sentiment for different
product characteristics.
48 Several utility prediction - - MAR 0.90 www.amazon.com
models are built using
multidimensional CART 0.97
adaptive regression,
classification and RandF 0.82
regression tree, random
forest approaches, Neural Net 0.75
neural network, and
deep neural network Deep NN 0.60
using two real-world
Amazon product
feedback datasets.
To reduce the time spent on choosing a popular new product, the authors have developed the
intelligent method of a competitive product choosing based on the emotional feedbacks coloring. The
proposed method is illustrated schematically (Fig.1) and is represented by the following steps.
Step 1. Connect the necessary libraries. They provide general mathematical and numerical
operations in the form of pre-compiled, fast functions. They are combined into high-level packages
(Block 1).
Step 2. Select the desired site for further work (Block 2).
Step 3. Enter the link to the product categories that user needs (Block 3).
Step 4. Parsing of user feedbacks to goods is performed. Divide them by product id and add to the
text file. A separate feedback for each line (Block 4).
Step 5. Creating a database of feedbacks collected together (Block 5).
Step 6. Connection of feedbacks from a DB (Block 5) for the further work (Block 6).
Step 7. The raw text is quite messy for these feedbacks, so seller need to clean the text to make it
easier for the algorithm to recognize the mood (Block 7).
Step 8. Vectorization is performed (Block 8). Vectorization is a type of program parallelization in
which single-threaded applications that perform one operation at a time are modified to perform
several operations of the same type at the same time. To perform the vectorization process seller need
to: tokenization (Block 8.1), ignoring single characters (Block 8.3), converting text into numerical
vectors and n-grams (Block 8.5) and then form an array of values (Block 8.6).
Data Mining Text Conversion Text Classification
Connection of 6
Input 10
feedbacks
of cleaned data
Connecting the 1 Cleaning and 7 Selection of ML classifier 11
necessary libraries processing
11.4
Selection of the 11.1 Test sample of
2 Vectorization of ML base 8 classified text
Site selection regularization
Process of “hot codding” parameter
Tokenization 8.1
Link to product 3 11.3
category 8.2
Test sample Cross-checking by classical methods of
8.3 Selection of 11.2
“vocabulary” Ignore of single ML classification:
classification method • Support Vector Classifier,
characters
JSON Parsing 4 • Stochastic Gradient Decent Classifier,
• Random Forest Classifier,
Parser structure: 8.5 • Decision Tree Classifier,
• Good Id Convert text to • Gaussian Naive Bayes,
numeric vector 11.5
• response txt • K-Neighbors Classifier,
8.4 representation of Construction of a
Test sample • Ada Boost Classifier,
“stop words” words and n-grams classification model • Logistic Regression.
DB of JSON5 Table forming
8.6
feedbacks Classification 12
Category positive: excellent, perfect, wonderful, amazing and wonderful.
Category negative: worst, waste, horrible, bad and boring.
DB 9 13
of converted text Result output
Figure 1: Structure of the intelligent method of a competitive product choosing based on the emotional feedbacks coloring
Step 8.1. Tokenization (Block 8.1) is required to replace a confidential data item with a non-
confidential equivalent called a token that has no independent meaning / significance for external or
internal use. Sample, which evaluates the quality of the constructed model. The quality assessment
made from the test sample can be used to select the best model (Block 8.2).
Step 8.2. The process of ignoring single characters that interfere with the program (Block 8.3) is
carried out on the basis of a ready set of “stop words” (Block 8.4). Stop words are very common
words such as “if”, “but”, “we”, “he”, “she” and “they”. Seller can usually delete these words without
changing the text semantics (but not always), improving the model performance.
Step 8.3. In this step, the text is converted into numerical vectors and add n-grams (Block 8.5). N-
grams are simply a sequence of n elements (sounds, syllables, words or symbols) that go in a row in a
text. If seller divide the text into several small fragments represented by N-grams, they are easy to
compare with each other and thus obtain a degree of similarity of the analyzed documents. N-grams
are often used successfully to categorize text and language. In addition, they can be used to create
functions that allow seller to gain knowledge from text data.
Step 8.4. Forming an array of vectorized text values for further storage in the database (Block 8.6).
Step 9. Forming a database of the converted text to select the classifier model (Block 9).
Step 10. Enter the cleared data (Block 10).
Step 11. The next step is to select the most optimal classifier by machine learning algorithms
(Block 11).
Step 11.1. Search for the best regularization parameter (Block 11.1). The regularization parameter
reduces retraining, which reduces the variance of the estimated regression parameters. Seller need to
use it. Cross-checking by eight classical classification methods: Support Vector Classifier, Stochastic
Gradient Decent Classifier, Random Forest Classifier, Decision Tree Classifier, Gaussian Naive
Bayes, K-Neighbors Classifier, Ada Boost Classifier, Logistic Regression (Block 11.3) [50, 52-54]. A
ready-made training sample (Block 11.4) was used to check the algorithms.
Step 11.2. Based on the obtained results, the best classification method is chosen (Block 11.2) of
the classification with the highest regularization parameters (Block 11.1) and form a classification
model (Block 11.5) based on the test sample (Block 11.4).
Step 12. Next is the data classification (Block 12) for positive and negative comments. Namely,
positive feedbacks are divided into categories: excellent, perfect, wonderful, amazing and wonderful.
Negatives are divided into categories: worst, waste, horrible, bad and boring.
Step 13. The last step is to derive the result on the basis of which seller can decide on a profitable
investment in a new product.
The results of experimental studies confirmed the correctness of the developed method, which will
be described in more detail in the camera-ready and the presentation.
4. Experimental results and Discussion
Python was chosen to conduct an intelligent method of choosing a competitive product based on
the feedback emotional coloring. The following libraries were used: pandas, numpy, train_test_split,
SVC, GridSearchCV, SGDClassifier, RandomForestClassifier, DecisionTreeClassifier, GaussianNB
KNeighborsClassifier, AdaBoostClassifier, LogisticRegression.
4477 feedbacks from the Rozetka site were used as input [51]. Feedbacks are collected by parsing
and formed into a JSON file, which presents the product id and feedback on it.
Next, the data was cleaned for the presence of characters that do not affect the content of the text:
“.”, “;”, “:”, “!”, “”, “?”, “,”, “"”, “()”, “[]”. The next step is removing the “stop words”, which are in
the Russian language, because most feedbacks are in Russian, further research will consider the
possibility of feedback distributing by language, for better results.
After vectorization and lemmatization, the classification of the most optimal classifiers of machine
learning algorithms was performed. Therefore, results will be cross-checked (Table 2) and evaluated
based on 8 different methods: Support Vector Classifier, Stochastic Gradient Decent Classifier,
Random Forest Classifier, Decision Tree Classifier, Gaussian Naive Bayes, K-Neighbors Classifier,
Ada Boost Classifier, Logistic Regression.
Table 2
The results of cross-evaluation
# Method Prediction assessment
1 SupportVectorClassifier 0.73
2 StochasticGradientDecentC 0.74
3 RandomForestClassifier 0.78
4 DecisionTreeClassifier 0.67
5 GaussianNB 0.77
6 KNeighborsClassifier 0.76
7 AdaBoostClassifier 0.76
8 LogisticRegression 0.73
Table 2 shows that all methods showed poor results, due to the fact that the feedbacks include both
Ukrainian and Russian languages. However, the best is RandomForestClassifier, with a forecast score
of 0.78, which is a valid score for evaluation. Probably, the model accuracy can be increased by
increasing the dataset size, as the four thousand dataset is quite modest. Also, it would be possible to
reduce the problem to a binary classification of feedbacks to positive and negative, which would also
increase accuracy.
Next, the feedbacks will be arranged into positive and negative based on the classification method
RandomForestClassifier (Fig. 4). Arranging is based on the inverted n-gram index and relative to the
selected product with id – 122360970 (one of the most popular headphones).
Table 3
Arranged feedbacks by product with id – 122360970
ID poswords pos_importance negwords neg_importance
122360970 top 22.57 the worst -0.77
122360970 good 1.82 horror -0.76
122360970 super 0.85 I do not recommend -0.76
122360970 cool 0.71 badly -0.69
122360970 best 0.71 nonsense -0.62
122360970 super-duper 0.71 broke -0.62
122360970 cool 0.58 bad -0.58
122360970 best 0.56 stopped working -0.56
122360970 excellent 0.55 negative -0.56
122360970 reliable 0.53 disappointed -0.54
When arranging (see table 3), 10 positive and negative words were separated relative to the n-gram
indices. Accordingly, for the product with id – 122360970, the following positive words with the
largest n-gram index were identified: “top” (2.57) and “good” (1.82). Also, negative words with the
lowest n-gram index – “worst” (-0.77), “horror” (-0.76). From the numerical characteristics of n-
grams, show that the positive words in the feedbacks more than negative, respectively, we can
conclude that the product with id – 122360970, is of good quality and worth choosing for resale.
Therefore, the developed method differs from analogues (see table 1) in that it will allow to parse
the relevant data from the target site, and the classification is tested by eight classical classification
methods, namely Support Vector Classifier, Stochastic Gradient Decent Classifier, Random Forest
Classifier, Decision Tree Classifier, Gaussian Naive Bayes, K-Neighbors Classifier, Ada Boost
Classifier, Logistic Regression. That gives the chance to make administrative decisions concerning
profitable investment in the new goods.
5. Conclusions
The intelligent method of a competitive product choosing based on the emotional feedbacks
coloring, based on which the seller can make management decisions regarding profitable investments
in a new product, and thus reduce the risks of non-profit sales. Also, the proposed method reduces the
time spent searching for popular and quality products based on user feedback.
The method was implemented on feedbacks (4477 feedbacks) from the Rozetka website. Feedback
is collected by parsing and formed into a JSON file and cleaned of unnecessary characters and “stop
words”. After vectorization, lemmatization and classification of the most optimal classifiers of
machine learning algorithms: Support Vector Classifier, Stochastic Gradient Decent Classifier,
Random Forest Classifier, Decision Tree Classifier, Gaussian Naive Bayes, K-Neighbors Classifier,
Ada Boost Classifier, Logistic Regression. All methods showed poor result, due to the fact that the
feedbacks include both Ukrainian and Russian feedbacks. However, the best is
RandomForestClassifier, with a forecast score of 0.78. Next, the feedbacks were sorted into positive
and negative based on the RandomForestClassifier classification. The words with the largest n-gram
index are highlighted: positive (“top” (2.57) and “good” (1.82)) and negative (“worst” (-0.77),
“horror” (-0.76)). The positive words in the feedbacks are more than the negative ones, based on the
n-gram indices, respectively, it can be concluded that they are of good quality and worth choosing for
resale.
Areas of further research include an in-depth study of the effectiveness of the developed method in
expanding the range of goods and their geography, as well as a significant increase in the number of
user feedbacks and the ability to distribute feedback by languages. In addition, it is worth trying to
reduce the problem to a binary classification of feedbacks to positive and negative, which would also
increase accuracy. Moreover, authors are going to explore ontology models [55] and Deep Learning
[56] in domains above.
6. References
[1] K. Kaushik, R. Mishra, N. Rana, Y. K. Dwivedi, Exploring reviews and review sequences on e-
commerce platform: A study of helpful reviews on Amazon.in. Journal of Retailing and
Consumer Services, 45 (2018) 21–32. https://doi:10.1016/j.jretconser.2018.08.002
[2] S. Saumya, J.P. Singh, & Y.K. Dwivedi, Predicting the helpfulness score of online reviews using
convolutional neural network. Soft Comput 24 (2020) 10989–11005.
https://doi.org/10.1007/s00500-019-03851-5
[3] Y. Zhang, Z. Lin, Predicting the helpfulness of online product reviews: A multilingual approach.
Electronic Commerce Research and Applications, 27 (2018) 1–10. doi:
10.1016/j.elerap.2017.10.008
[4] J. Wu, Y. Wu, J. Sun, & Z. Yang, User reviews and uncertainty assessment: A two stage model
of consumers' willingness-to-pay in online markets. Decision Support Systems, 55.1 (2013) 175-
185.
[5] J. P. Singh, S. Irani, N. P. Rana, Y. K. Dwivedi, S. Saumya, & P. K. Roy, Predicting the
“helpfulness” of online consumer reviews. Journal of Business Research, 70 (2017) 346-355.
[6] L. Celardo, & M. G. Everett, Network text analysis: A two-way classification approach.
International Journal of Information Management, 51 (2020) 102009.
[7] M. Syamala, & N. J. Nalini, A Filter Based Improved Decision Tree Sentiment Classification
Model for Real-Time Amazon Product Review Data. International Journal of Intelligent
Engineering and Systems, 13.1 (2020) 191-202.
[8] M. A. Mirtalaie, & O. K. Hussain, Sentiment aggregation of targeted features by capturing their
dependencies: Making sense from customer reviews. International Journal of Information
Management, 53 (2020) 102097.
[9] Y. Liu, C. Jiang, & H. Zhao, Using contextual features and multi-view ensemble learning in
product defect identification from online discussion forums. Decision Support Systems, 105
(2018) 1-12.
[10] B. Bansal, S. Srivastava, Hybrid attribute based sentiment classification of online reviews for
consumer intelligence. Appl Intell, 49 (2019) 137–149. https://doi.org/10.1007/s10489-018-
1299-7
[11] C. Guo, Z. Du, & X. Kou, Products Ranking Through Aspect-Based Sentiment Analysis of
Online Heterogeneous Reviews. J. Syst. Sci. Syst. Eng. 27 (2018) 542–558.
https://doi.org/10.1007/s11518-018-5388-2
[12] J.-W. Bi, Y. Liu, Z.-P. Fan, Representing sentiment analysis results of online reviews using
interval type-2 fuzzy numbers and its application to product ranking. Information Sciences, 504
(2019) 293–307. doi:10.1016/j.ins.2019.07.025
[13] L. Zheng, Z. He, & S. He, A novel probabilistic graphic model to detect product defects from
social media data. Decision Support Systems, 137 (2020) 113369. doi:10.1016/j.dss.2020.113369
[14] Y. Liu, C. Jiang, & H. Zhao, Assessing product competitive advantages from the perspective of
customers by mining user-generated content on social media. Decision Support Systems, 123
(2019) 113079. https://doi.org/10.1016/j.dss.2019.113079
[15] S. Khedkar, S. Shinde, Ensemble Classifier for Praise or Complaint Classification and
Visualization from Big Data. In: Dey N., Mahalle P., Shafi P., Kimabahune V., Hassanien A.
(eds) Internet of Things, Smart Computing and Technology: A Roadmap Ahead. Studies in
Systems, Decision and Control, vol. 266, (2020) 97-118. https://doi.org/10.1007/978-3-030-
39047-1_5
[16] M. Savargiv, B. Masoumi, & M.R. Keyvanpour, A new ensemble learning method based on
learning automata. J Ambient Intell Human Comput. (2020) https://doi.org/10.1007/s12652-020-
01882-7
[17] S. Khedkar, S. Shinde, Deep Learning-Based Approach to Classify Praises or Complaints from
Customer Reviews. In: Bhalla S., Kwan P., Bedekar M., Phalnikar R., Sirsikar S. (eds)
Proceeding of International Conference on Computational Science and Applications. Algorithms
for Intelligent Systems. Springer, Singapore, (2020) 391-402. https://doi.org/10.1007/978-981-
15-0790-8_38
[18] Z.-P. Fan, G.-M. Li, Y. Liu, Processes and methods of information fusion for ranking products
based on online reviews: An overview. Information Fusion, (2020) 1-27.
doi:10.1016/j.inffus.2020.02.007.
[19] S. Saumya, J. P. Singh, A. M. Baabdullah, N. P. Rana, & Y. K. Dwivedi, Ranking online
consumer reviews. Electronic Commerce Research and Applications, 29 (2018) 78-89.
[20] H. C. Lee, H. C. Rim, & D. G. Lee, Learning to rank products based on online product reviews
using a hierarchical deep neural network. Electronic Commerce Research and Applications, 36
(2019) 100874.
[21] C. Vo, D. Duong, D. Nguyen, and T. Cao, From Helpfulness Prediction to Helpful Review
Retrieval for Online Product Reviews. In Proceedings of the Ninth International Symposium on
Information and Communication Technology (SoICT 2018). Association for Computing
Machinery, New York, NY, USA, (2018) 38–45. https://doi.org/10.1145/3287921.3287931
[22] J. Du, J. Rong, S. Michalska, H. Wang, Y. Zhang, Feature selection for helpfulness prediction of
online product reviews: An empirical study. PLoS ONE, 14.12 (2019) e0226902.
https://doi.org/10.1371/journal.pone.0226902
[23] M. Bilal, M. Marjani, I. A. T. Hashem, A. M. Abdullahi, M. Tayyab and A. Gani, Predicting
Helpfulness of Crowd-Sourced Reviews: A Survey. In Proceedings of the 2019 13th
International Conference on Mathematics, Actuarial Science, Computer Science and Statistics
(MACS), Karachi, Pakistan, (2019) pp. 1-8, doi: 10.1109/MACS48846.2019.9024814.
[24] X. Luo, & Y. Yi, Topic-Specific Emotion Mining Model for Online Comments. Future Internet,
11.3 (2019) 79.
[25] S. Mitra, & M. Jenamani, SentiCon: A Concept Based Feature Set For Sentiment Analysis. In
Proceedings of the 2018 IEEE 13th International Conference on Industrial and Information
Systems (ICIIS) (2018) 246-250.
[26] Z. Zeng, Z. Zhou, and X. Mu, User review helpfulness assessment based on sentiment analysis,
The Electronic Library, 38.2 (2020) 337-351. https://doi.org/10.1108/EL-08-2019-0200
[27] F. Fouladfar, M. N. Dehkordi and M. E. Basiri, Predicting the Helpfulness Score of Product
Reviews Using an Evidential Score Fusion Method, IEEE Access, 8 (2020) 82662-82687. doi:
10.1109/ACCESS.2020.2988872.
[28] M. E. Basiri, & S. Habibi, Review Helpfulness Prediction Using Convolutional Neural Networks
and Gated Recurrent Units. Proceedings of the 2020 6th International Conference on Web
Research (ICWR) (2020) 191-196. doi: 10.1109/icwr49608.2020.9122297
[29] L. Kong, C. Li, J. Ge, V. Ng, & B. Luo, Predicting Product Review Helpfulness A Hybrid
Method. IEEE Transactions on Services Computing, (2020). doi: 10.1109/TSC.2020.3041095.
[30] H. Dong, Y. Hou, M. Hao, J. Wang, & S. Li, Method for Ranking the Helpfulness of Online
Reviews Based on SO-ILES TODIM. IEEE Access, 9 (2020) 1723-1736. Doi:
10.1109/ACCESS.2020.3040151
[31] M.S.I. Malik, A. Hussain, Helpfulness of product reviews as a function of discrete positive and
negative emotions. Computers in Human Behavior, 73 (2017) 290–302. doi:
10.1016/j.chb.2017.03.053
[32] L. Li, T.T. Goh, & D. Jin, How textual quality of online reviews affect classification
performance: a case of deep learning sentiment analysis. Neural Comput & Applic, 32 (2020)
4387–4415. https://doi.org/10.1007/s00521-018-3865-7
[33] S. P. Eslami, M. Ghasemaghaei, & K. Hassanein, Which online reviews do consumers find most
helpful? A multi-method investigation. Decision Support Systems, 113 (2018) 32-42.
[34] U. Chakraborty, and S. Bhat, Credibility of online reviews and its impact on brand image,
Management Research Review, 41.1 (2018) 148-164. https://doi.org/10.1108/MRR-06-2017-
0173
[35] R. Liang, Jq. Wang, A Linguistic Intuitionistic Cloud Decision Support Model with Sentiment
Analysis for Product Selection in E-commerce. Int. J. Fuzzy Syst., 21 (2019) 963–977.
https://doi.org/10.1007/s40815-019-00606-0
[36] B. Samal, A. K. Behera, & M. Panda, Performance analysis of supervised machine learning
techniques for sentiment analysis. On Proceedings of the 2017 Third International Conference on
Sensing, Signal Processing and Security (ICSSS) (2017) 128-133.
doi:10.1109/ssps.2017.8071579
[37] X. Sun, M. Han, J. Feng, Helpfulness of online reviews: Examining review informativeness and
classification thresholds by search products and experience products. Decision Support Systems,
(2019) 113099. doi: 10.1016/j.dss.2019.113099
[38] M. Akbarabadi, & M. Hosseini, Predicting the helpfulness of online customer reviews: The role
of title features. International Journal of Market Research, 62.3 (2020) 272-287.
https://doi.org/10.1177/1470785318819979
[39] P. Wu, X. Li, S. Shen, & D. He, Social media opinion summarization using emotion cognition
and convolutional neural networks. International Journal of Information Management, 51 (2020)
101978.
[40] M. Arif, U. Qamar, F.H. Khan, S. Bashir, A Survey of Customer Review Helpfulness Prediction
Techniques. In: Arai K., Kapoor S., Bhatia R. (eds) Intelligent Systems and Applications.
IntelliSys 2018. Advances in Intelligent Systems and Computing, vol 868. Springer, Cham,
(2019) 215-226. https://doi.org/10.1007/978-3-030-01054-6_15
[41] E. Kauffmann, J. Peral, D. Gil, A. Ferrández, R. Sellers, H. Mora, Managing Marketing
Decision-Making with Sentiment Analysis: An Evaluation of the Main Product Features Using
Text Data Mining. Sustainability, 11.15 (2019) 4235. doi:10.3390/su11154235
[42] S. Saumya, J. Kumar, J.P. Singh, Genre Fraction Detection of a Movie Using Text Mining. In:
Chaki R., Cortesi A., Saeed K., Chaki N. (eds) Advanced Computing and Systems for Security.
Advances in Intelligent Systems and Computing, vol 666 (2018) 167-177. Springer, Singapore.
https://doi.org/10.1007/978-981-10-8180-4_11.
[43] J.P. Singh, N.P. Rana, W. Alkhowaiter, Sentiment Analysis of Products’ Reviews Containing
English and Hindi Texts. In: Janssen M. et al. (eds) Open and Big Data Management and
Innovation. I3E 2015. Lecture Notes in Computer Science, vol 9373. Springer, Cham (2015)
416-422. https://doi.org/10.1007/978-3-319-25013-7_33
[44] V. Kumar, S. Pasari, V. P. Patil, S. Seniaray, Machine Learning based Language Modelling of
Code Switched Data. In Proceedings of the 2020 International Conference on Electronics and
Sustainable Communication Systems (ICESC) (2020) 552–557.
doi:10.1109/ICESC48915.2020.9155695
[45] V. Battu, V. Batchu, R. R. R. Gangula, M. M. K. R. Dakannagari, & R. Mamidi, Predicting the
Genre and Rating of a Movie Based on its Synopsis. In PACLIC. (2018) 52-62.
[46] J. Wehrmann, W. E. Becker, R. C. Barros, A multi-task neural network for multilingual
sentiment classification and language detection on Twitter. In Proceedings of the 33rd Annual
ACM Symposium on Applied Computing – SAC'18 (2018) 1805–1812.
doi:10.1145/3167132.3167325
[47] W.-T. Chu and H.-J. Guo, Movie Genre Classification based on Poster Images with Deep Neural
Networks. In Proceedings of the Workshop on Multimodal Understanding of Social, Affective
and Subjective Attributes (MUSA2 '17). Association for Computing Machinery, New York, NY,
USA, (2017) 39–45. https://doi.org/10.1145/3132515.3132516
[48] M.S.I. Malik, Predicting users’ review helpfulness: the role of significant review and reviewer
characteristics. Soft Comput, 24 (2020) 13913–13928. https://doi.org/10.1007/s00500-020-
04767-1
[49] S. Zhou, & B. Guo, The order effect on online review helpfulness: A social influence
perspective. Decision Support Systems, 93 (2017) 77-87.
[50] H. Lipyanina, V. Maksymovych, A. Sachenko, T. Lendyuk, A. Fomenko, I. Kit, Assessing the
Investment Risk of Virtual IT Company Based on Machine Learning. In: Babichev S., Peleshko
D., Vynokurova O. (eds) Data Stream Mining & Processing. DSMP 2020. Communications in
Computer and Information Science, vol 1158 (2020) 167-187 Springer, Cham.
https://doi.org/10.1007/978-3-030-61656-4_11
[51] Internet-Shop “Rozetka™” – https://rozetka.com.ua/ua/
[52] C. Wang, N. Shakhovska, A. Sachenko, M. Komar, A New Approach for Missing Data
Imputation in Big Data Interface. Information Technology and Control, 49.4, (2020) 541-555.
https://doi.org/10.5755/j01.itc.49.4.27386.
[53] H. Lipyanina, S. Sachenko, T. Lendyuk, A. Sachenko, Targeting Model of HEI Video Marketing
based on Classification Tree. In: Proceedings of the 16th International Conference on ICT in
Education, Research and Industrial Applications. Kharkiv, Ukraine, October 6-10, 2020, pp. 487-
498. ISSN: 1613-0073. http://ceur-ws.org/Vol-2732/20200487.pdf.
[54] H. Lipyanina, A. Sachenko, T. Lendyuk, S. Nadvynychny, S. Grodskyi, Decision Tree Based
Targeting Model of Customer Interaction with Business Page. In: Proceedings of the third
International Workshop on Computer Modeling and Intelligent Systems (CMIS-2020), CEUR
Workshop Proceedings, vol. 2608, 2020, pp. 1001-1012. ISSN: 1613-0073. Electronic copy at:
http://ceur-ws.org/Vol-2608/paper75.pdf.
[55] C. Shu, D. Dosyn, V. Lytvyn, V. Vysotska, A. Sachenko, S. Jun, Building of the Predicate
Recognition System for the NLP Ontology Learning Module. In: Proceedings of the 10th IEEE
International Conference on Intelligent Data Acquisition and Advanced Computing Systems:
Technology and Applications (IDAACS’2019), 18-21 September, 2019, Metz, France, vol. 2,
(2019) 802-808. https://doi.org/10.1109/IDAACS.2019.8924410.
[56] V. Golovko, S. Bezobrazov, A. Kroshchanka, A. Sachenko, M. Komar, A. Karachka,
Convolutional Neural Network Based Solar Photovoltaic Panel Detection in Satellite Photos. In:
Proceedings of the 9th IEEE International Conference on Intelligent Data Acquisition and
Advanced Computing Systems: Technology and Applications (IDAACS’2017), September 21-
23, 2017, Bucharest, Romania, (2017) 14-19. https://doi.org/10.1109/IDAACS.2017.8094501.