Scientific Journal of Informatics

Vol. 10, No. 3, Aug 2023

p-ISSN 2407-7658 http://journal.unnes.ac.id/nju/index.php/sji e-ISSN 2460-0040

Coastal Sentiment Review Using Naïve Bayes with Feature Selection

Genetic Algorithm

Oman Somantri1*, Ratih Hafsarah Maharrani2 , Santi Purwaningrum3

1,2
Cybersecurity Engineering Department, Department of Informatics,
State Polytechnic of Cilacap, Indonesia
3
Informatics Department, Department of Informatics,
State Polytechnic of Cilacap, Indonesia

Abstract.
Purpose: The tourism potential in the maritime sector can be Indonesia's mainstay at this time, especially in enjoying
the charm of the natural beauty of the coast as people know Indonesia is an archipelagic country. The purpose of this
study is to find the best model by applying the feature selection genetic algorithm (GA) and Information Gain (IG) to
get the best Naïve Bayes (NB) model and the best features to produce the best level of sentiment classification accuracy.
Methods: The stages of the research were carried out by going through the process of searching, pre-processing,
analyzing research data using the Naïve Bayes model and optimizing genetic algorithms, validating data, and model
evaluation.
Results: The experimental results show that the best model is naïve Bayes based on information gain and the genetic
algorithm yields an accuracy rate of 86.34%.
Novelty: The main contribution to this research is proposing a new model of the best NB optimization model by
applying an optimization algorithm in the search for feature selection to increase sentiment classification accuracy.
Keywords: Coastal, Naïve bayes, Information gain, Feature selection, Genetic algorithm
Received April 2023 / Revised April 2023 / Accepted May 2023

This work is licensed under a Creative Commons Attribution 4.0 International License.

INTRODUCTION
Indonesia has currently named an archipelagic country that is famous around the world for its biggest
potential in the maritime sector. This maritime sector has become one of the prima donnas, both local and
foreign tourists deliberately visit to enjoy the natural beauty of Indonesia's coasts. Assessment of people's
perspectives on such a beautiful coastal destination will be very influential on the number of people who
are interested in following others that have previously seen or enjoyed the beach [1], [2]. Social media is
often used by people or tourists in assessing the beauty of beaches [3],[4]. We cannot deny that currently,
social media has a big influence compared to other media such as newspapers and other print media [5],
[6].

Social media is media that provides an overview of something even if the news is true or not. It is very
influential on a person's life and decisions in taking action [7]. Assessment of someone's review of a place
is currently often carried out by other people through social media as a means to give a review or assessment
of the place therefore others know regardless of whether the review is positive or negative [8]. The same is
true with many people reviewing coastal locations, especially in southern coastal areas of Java. The
assessment of coastal sentiment reviews that many people do indirectly greatly influences the potential for
coastal maritime tourism.

Sentiment analysis is a field of science that utilizes artificial intelligence to enable it to provide decision
support in assessing sentiment by categorizing whether the sentiment is positive, negative, or neutral [9].
The application of SA is often carried out using machine learning in which each applied algorithm produces
a different level of accuracy according to the strengths and weaknesses of each model. Various fields have
used sentiment analysis as a model [10], [11] with the hope that it can help in providing decision support
for each policy to be decided. Some examples of research in the field of sentiment analysis research are

*
Corresponding author.
Email addresses: [email protected] (Somantri)
DOI: 10.15294/sji.v10i3.43988

Scientific Journal of Informatics, Vol. 10, No. 3, Aug 2023 | 229

applied to film reviews [12], [13], application reviews on the Google Play Store [14], hotel reviews [15],
[16], restaurant customer review [17], and product reviews [18]–[20].

Of the several algorithms often used according to their advantages for sentiment analysis, especially
classification, are the neural network algorithm and the naïve Bayes (NB) algorithm. Neural networks have
the advantage of being able to carry out learning to work based on the initial experience of the model and
can perform calculations in parallel [21], [22]. Naïve Bayes is usually used to apply small-scale data for
training, besides has been widely used for data processing, especially text mining since it has a better level
of accuracy [23], [24]. Based on its advantages, NB is applied for sentiment reviews of coastal assessments
in the hope of producing a good and high level of model accuracy.

Figure 1. Map of the southern coast of Java Island, Indonesia

The problem that occurs in this algorithm is that there are still parameter values that must be given manually,
making it difficult to get the best model and the best weight features. In addition, the determination of the
initial weight value in the processed model is still not optimal, so it requires optimization. The purpose of
this study is to find the best model by applying an optimization algorithm to get the best model of the two
algorithms applied and to get the best weight and parameter values to produce the best level of sentiment
classification accuracy.

The research was conducted by Yan, Yingwei, et al [25] using social media as material in assisting planning
for the recovery of tourist destinations, especially in the Lombok and Bali areas after the 2018 disaster. The
public's view of post-disaster tourist destinations, especially in the Bali and Lombok regions, relies on beach
tourism. This study proposes the Latent Dirichlet Allocation (LDA) method with the data from Twitter, and
the results of the research show that the proposed and implemented approach can effectively reveal various
kinds of sentiments and community perspectives on issues regarding post-disaster tourism recovery from
time to time.

Subsequent research was conducted by Park, Eunhye., et al [26] to empirically test the effect of news in
predicting the level of tourist arrivals. In this study, data sources originating from news source topics were
extracted into data used for forecasting tourist arrivals, especially in Hong Kong. The proposed method is
the Autoregressive Integrated Moving Average (ARIMA) method by performing feature selection for
selecting variables first. The proposed research model helps tourist destinations in overcoming the
externalities of reporting in the news media that affect people's sentiments in assessing a tourist destination.
Another study related to destination sentiment was conducted by Ali, and Twiland [27] to get the best model
of tourist experience sentiment in Morocco. In this study, a combined model is proposed using a
combination of topic modeling and lexicon-based algorithms using Latent Dirichlet Allocation (LDA)
where the data comes from TripAdvisor reviews of various tourist attractions in Marrakech, Morocco. The
next research is slightly different, which was conducted by Sohrabi, B [28] that proposed a model for

230 | Scientific Journal of Informatics, Vol. 10, No. 3, Aug 2023

predicting tourist destination visits based on comments and interests using text mining by the X-means
clustering model and classification with a Decision Tree.

Based on some of the results of previous studies, the researchers conducted experiments using an
algorithmic model without optimizing the model, therefore the resulting level of accuracy was not optimal.
On the other hand, these limitations are accompanied by the source of the research dataset used in the form
of a review of tourist attractions. It use social media data and its influence has an impact on the pre-
processing of data that leads to different resulting level of accuracy. For this reason, one of the efforts to
increase the level of accuracy produced by the sentiment review model of the coast as a tourist attraction is
to propose a new model using Naïve Bayes (NB). This article proposes Naïve Bayes for the classification
of sentiment review destinations on the southern coast of Java Island as a recommendation to increase
maritime tourism visits, especially beach tourism based on feature selection using a genetic algorithm. The
main contribution to this study is to apply an optimization algorithm for feature selection in the NB model
to increase the accuracy of the sentiment review classification.

METHODS
Dataset
The research carried out is using experimental research methods, where the data will be processed and input
into the model to get the model with the best level of accuracy. The data used in this study were taken from
the website https://www.google.com/maps, by entering the keyword "beach" which then the search results
contained a review of places according to the desired beach name with different star rating values. The
difference is between 1 to 5 where the rating value of 5 is the highest positive sentiment. The data taken is
Indonesian language text data taken from 2018 to 2021 which is then processed into the desired model of
390 data, examples of data taken are shown in Table 1.

Table 1. Example of a research dataset used in Bahasa

Sentiment Positive Sentiment Negative
• [Pantai teluk penyu Cilacap, tempat yang bagus • [Pantainya jorok, tidak ada tempat sampah,
untuk melihat luasnya lautan dan mendengarkan masyarakat dan pengelola pantai harus sadar
deburan ombak laut. Datang ke tempat ini pas bukan bebersih.]
waktu liburan jadi lumayan agak lenggang, banyak
penjual kuliner disekitar tempat ini, juga ada
beberapa tempat untuk berteduh para pengunjung
dari teriknya matahari.]

• [Pantai yg ramah buat ciblonan.. walau agak kotor • [Pantai nya kotor banget, beli mendoan seporsi
tapi masih tetap indah. Kita juga bisa menyebrang isi 8 biji 28.000, wkwk padahal di pantai
naik perahu menuju pantai pasir putih, banyak batu2 sodong mendoan seporsi isi 8 cuma 15.000.]
kecil yg indah berwarna warni.]

The process of determining to label in this study was carried out based on the assessment of the asterisk. It
is included in negative sentiment if the data is given 1-3 stars, and it is included in the positive sentiment
category if the data is rated 5 stars. In this study, the sentiment sought in the model is only limited to positive
and negative sentiments.
Preprocessing Data
In the proposed research, the next step after the dataset in the form of a coastal review text is obtained is to
do data preprocessing. The stages of the preprocessing process in this study were carried out to obtain the
expected text data, namely data cleansing, tokenization, stemming, and data filtering [29], [30]. The next
process is the data is input into a predetermined model, namely the model using Naïve Bayes.

Scientific Journal of Informatics, Vol. 10, No.2, Aug 2023 | 231

 Figure 2. The proposed method framework
The next stage is to carry out model optimization using the feature selection algorithm and conduct data
validation using the cross-validation method. It is expected that there will be an increase in the accuracy of
sentiment classification and it will get better [31]. The process of the stages carried out in this study can be
seen in Figure 2.

Naïve Bayes Algorithm

Naïve Bayes is a method that is included in a classification algorithm derived from the concept of processing
statistical data and probabilities that can predict these opportunities based on previous experience [32], [33].
The NB equation is shown in equation (1) [34].

(1)

In equation (1) where X is data with unknown class, H is hypothesis data X, P(H|X) is the probability of
hypothesis H based on X conditions, P(H) is the hypothesis probability of H, P(X |H) is the probability of
X based on the conditions in the hypothesis H, and P(X) is the probability of X.

To get a performance value from the sentiment classification obtained, this study uses equation (2) [35],
[36].
𝑇𝑃+𝑇𝑁
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 = (2)
𝑇𝑃+𝑇𝑁+𝐹𝑃+𝐹𝑁

In equation (2), where TP is True Positive, TN is True Negative, FP is False Positive, and FN is False
Negative.

RESULTS AND DISCUSSIONS

Experimental Results Using Naïve Bayes
The first stage is to experiment with several methods using a predetermined algorithm that is proposed to
obtain an expected model. The Naïve Bayes algorithm is applied to the preliminary experiment, in this
process NB is expected to give the best accuracy rate of the proposed model. The results of applying the
NB model to the experiment are shown in Table 2.

232 | Scientific Journal of Informatics, Vol. 10, No. 3, Aug 2023

 Table 2. Experimental Results Using Naïve Bayes
Accuracy
Fold
Stratified Sampling (%) Shuffled Sampling (%) Linear Sampling (%)
10 61.79 62.31 55.90
9 63.34 61.83 55.41
8 63.87 62.08 54.11
7 64.10 62.80 54.61
6 67.44 61.54 53.08
5 62.05 64.10 51.03
4 63.57 64.62 51.35
3 67.44 63.33 50.26
2 65.13 63.85 42.56

As shown in Table 2, it can be seen that the highest level of accuracy obtained was 67.44% which was
obtained using the Fold=6 parameter with the stratified sampling method. In this experiment, the results
still need improvements therefore at a later stage another model was implemented for optimization.

Naïve Bayes Experiment Based on Information Gain (IG)

The next stage of the process is to optimize the model to increase the accuracy of the coastal review
sentiment classification. At this stage, an optimization process is carried out using the Information Gain
(IG) method [37], [38], so that an increase in accuracy is obtained and the results can be seen in Table 2
and Figure 2.
Based on Figure 3 and Table 3, it can be seen that there is a significant change. The lowest accuracy level
produced is 58.72% by using fold = 2 and linear sampling. The highest level of accuracy in this model
changes quite high, this change is an effort that has been made, namely by applying Information Gain to
the Naïve Bayes algorithm so that there is an increase in the accuracy level of 80.91%. This change occurs
by using the parameter fold=9 and using linear sampling.

Table 3. Naive Bayes model experiment and information gain

Accuracy
Fold
Stratified Sampling (%) Shuffled Sampling (%) Linear Sampling (%)
10 73.08 72.05 79.49
9 74.88 71.49 80.91
8 75.89 71.82 79.76
7 70.54 75.16 75.10
6 71.54 74.36 76.92
5 72.31 70.26 73.33
4 64.84 75.89 61.36
3 74.87 69.49 66.15
2 64.62 67.18 58.72

90,00%
80,91%
80,00%
Acuracy

70,00% 67,44%

60,00%
Model

NB Stratified Sampling NB + IG Linear Sampling

Figure 3. Comparison of the application of the NB and NB+IG methods

Scientific Journal of Informatics, Vol. 10, No.2, Aug 2023 | 233

Optimization of Feature Selection
Efforts are being made to obtain a coastal sentiment review classification model. One other effort is to apply
an algorithm for feature selection, namely by using the genetic algorithm (GA) method. The genetic
algorithm is one of the optimization algorithms with its ability to be used for classification. It is used as a
feature selection to select the best features in the NB model so that the resulting level of accuracy is better
[39].
Based on the experimental results, a significant result was obtained compared to the previous one, shown
in Table 4 to produce the best model by determining the parameter values in the GA algorithm, namely
population = 5 and selection scheme = tournament.
Table 4. Experimental results NB+IG and GA, tournament.
Crossover type Fold Sampling Accuracy
uniform 9 linear 86.34%
uniform 10 linear 81.54%
uniform 9 shuffled 77.13%
uniform 10 shuffled 76.92%
uniform 9 stratified 77.17%
uniform 10 stratified 77.18%
shuffled 9 linear 85.29%
Shuffled 10 linear 85.64%
Shuffled 9 shuffled 76.18%
Shuffled 10 shuffled 75.90%
Shuffled 9 stratified 76.17%
Shuffled 10 stratified 75.90%

Table 5. Experimental results NB+IG and GA, roulette wheel

crossover type Fold Sampling Accuracy
uniform 9 linear 80.78%
uniform 10 linear 83.33%
uniform 9 shuffled 75.88%
uniform 10 shuffled 76.15%
uniform 9 Stratified 76.14%
uniform 10 Stratified 75.64%
shuffled 9 linear 76.11%
shuffled 10 linear 85.64%
shuffled 9 shuffled 75.92%
shuffled 10 shuffled 76.67%
shuffled 9 Stratified 76.15%
shuffled 10 Stratified 76.41%

To compare feature selection optimization, another experiment was carried out using different parameters,
namely population = 5 and selection scheme = roulette wheel which produced experimental results as
shown in Table 5.
The best model for optimizing the NB method using a genetic algorithm-based feature selection is 86.34%
with a micro average accuracy of 86.41%. This increase in accuracy has a significant impact on model
accuracy, starting from 80.91% to an increase of 5.43%. The experimental results obtained are shown in
Table 6, besides that by using formula (2) the micro average accuracy is 86.41%.
Table 6. Confusion matrix result
True Negative True Positive Class Precision
Prediction negative 100 13 88,50%
Prediction Positive 40 237 85,56%

𝑇𝑃 + 𝑇𝑁
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 =
𝑇𝑃 + 𝑇𝑁 + 𝐹𝑃 + 𝐹𝑁

234 | Scientific Journal of Informatics, Vol. 10, No. 3, Aug 2023

 237 + 100
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 =
237 + 100 + 40 + 13

337
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 = = 0,8641
390

To evaluate the performance evaluation of the model obtained, the AUC (Area Under the Curve) value is
applied [40]. In this best model, it was found that the AUC value was 0.79 and the AUC value for this
model is included in the "fair classification" criteria in the table. The AUC category table itself is shown in
Table 7.

Table 7. AUC Value, its meaning, and symbols

AUC Range Classification level
0,9 – 1.0 Excellent classification
0,8 – 0,9 Good classification
0,7 – 0.8 Fair classification
0,6 – 0,7 Poor classification
0,5 – 0,6 Failure

Figure 4. AUC value of the proposed model

Although the model proposed and obtained using feature selection-based Naïve Bayes is currently in the
"fair" category, in terms of accuracy it still produces a fairly high value of 86.35%. However, in the future,
it still requires an increase in accuracy, especially the resulting AUC value.

Evaluation Model
The results of the experiments that have been carried out show different levels of accuracy. This provides
evidence that the level of accuracy obtained does not depend on just one model or algorithm that can be
applied, but still requires optimization efforts that can be maximized if the desired model is not sufficient
or in accordance. Changes in the parameter values in each model greatly influence the level of accuracy
and this makes a great effort in determining the parameter values to match what we expect.
Based on the experimental results, the evaluation of the model was carried out by comparing several
experimental methods that obtained the results obtained using classic Naïve Bayes (NB), NB with
Information Gain, and NB and Information Gain optimized using a genetic algorithm (GA). The results of
the model evaluation are shown in Figure 5 and Table 8.

Scientific Journal of Informatics, Vol. 10, No.2, Aug 2023 | 235

 Table 8. Evaluate the naïve Bayes model
No. Model Akurasi
1 Naïve Bayes 67,44%
2 Naïve Bayes + IG 80,91%
3 NB + IG + GA - tournament 86,34%
4 NB + IG + GA - roulete wheel 85,64%

100,00%
86,34% 85,64%
80,91%
80,00% 67,44%
Accuracy

60,00%

40,00%

20,00%

0,00%
Model

Naïve Bayes Naïve Bayes + IG

NB + IG + GA - tournament NB + IG + GA - roulete wheel

Figure 5. Coastal review sentiment model evaluation chart

Based on Table 6, it can be seen that if we evaluate and compare the several models that have been obtained,
it can be seen that the highest accuracy rate is 86.34% using the NB_IG algorithm based on feature selection
using GA using selection scheme = tournament, and the folds used are 9, and population = 5. Based on the
results obtained in this case, the proposed model is a model that has a greater degree of accuracy compared
to other models.

CONCLUSION
The sentiment review assessment model for the coast using the Naïve Bayes algorithm has been obtained
after optimization with the highest accuracy rate of 86.34%. The model obtained provides a benefit that can
be used by policymakers, especially related parties, to improve coastal maritime tourism, be it services,
facilities, or other things that can be optimized. The level of accuracy produced at this time requires efforts
to improve accuracy, so further research are needed. It is recommended for further research to optimize
from various angles such as data pre-processing, selectinging the best parameters, and optimizing weight
values. In addition, it is necessary to apply other algorithms to obtain other experimental results therefore
the best model can be seen and applied.

ACKNOWLEDGEMENT
Thank you to the Ministry of Education and Culture through the Academic Directorate of Vocational
Education, and the Directorate General of Vocational Education, for funding this research through the
Beginner Lecturer Research scheme for the 2022 implementation year.

REFERENCES
[1] V. Teles da Mota, C. Pickering, and A. Chauvenet, “Popularity of Australian beaches: Insights from
social media images for coastal management,” Ocean Coast. Manag., vol. 217, p. 106018, Feb.
2022, doi: 10.1016/j.ocecoaman.2021.106018.
[2] I. I. Ibrahim, “Social Media Analysis in Building Customer Trust - Systematic Literature Review,”
in Proceedings of 2022 International Conference on Information Management and Technology,
ICIMTech 2022, Aug. 2022, pp. 39–44, doi: 10.1109/ICIMTech55957.2022.9915099.
[3] M. T. Cuomo, I. Colosimo, L. R. Celsi, R. Ferulano, G. Festa, and M. La Rocca, “Enhacing

236 | Scientific Journal of Informatics, Vol. 10, No. 3, Aug 2023

 Traveller Experience in Mobility Services Via Big Social Data Analytics,” Technol. Forecast. Soc.
Change, vol. 176, p. 121460, Mar. 2022, doi: 10.1016/j.techfore.2021.121460.
[4] N. Pruksorranan, “Conceptual Perspectives on Tourist Behavior toward Information Technology
and Social Media in Bangsaen Beach, Chonburi, Thailand,” in 2020 Joint International Conference
on Digital Arts, Media and Technology with ECTI Northern Section Conference on Electrical,
Electronics, Computer and Telecommunications Engineering (ECTI DAMT & NCON), Mar. 2020,
pp. 58–60, doi: 10.1109/ECTIDAMTNCON48261.2020.9090748.
[5] F. Mirzaalian and E. Halpenny, “Exploring destination loyalty: Application of social media
analytics in a nature-based tourism setting,” J. Destin. Mark. Manag., vol. 20, p. 100598, Jun. 2021,
doi: 10.1016/j.jdmm.2021.100598.
[6] D. Obembe, O. Kolade, F. Obembe, A. Owoseni, and O. Mafimisebi, “Covid-19 and the tourism
industry: An early stage sentiment analysis of the impact of social media and stakeholder
communication,” Int. J. Inf. Manag. Data Insights, vol. 1, no. 2, p. 100040, Nov. 2021, doi:
10.1016/j.jjimei.2021.100040.
[7] R. Dolan, Y. Seo, and J. Kemper, “Complaining practices on social media in tourism: A value co-
creation and co-destruction perspective,” Tour. Manag., vol. 73, pp. 35–45, Aug. 2019, doi:
10.1016/j.tourman.2019.01.017.
[8] W. He, G. Yan, J. Shen, and X. Tian, “Developing a workflow approach for mining online social
media data,” in 2017 IEEE SmartWorld Ubiquitous Intelligence and Computing, Advanced and
Trusted Computed, Scalable Computing and Communications, Cloud and Big Data Computing,
Internet of People and Smart City Innovation,
SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI 2017 - Conference Proceedings, Aug. 2018,
pp. 1–6, doi: 10.1109/UIC-ATC.2017.8397648.
[9] M. B. Nasreen Taj and G. S. Girisha, “Insights of strength and weakness of evolving methodologies
of sentiment analysis,” Glob. Transitions Proc., vol. 2, no. 2, pp. 157–162, Nov. 2021, doi:
10.1016/j.gltp.2021.08.059.
[10] M. V. Mäntylä, D. Graziotin, and M. Kuutila, “The evolution of sentiment analysis—A review of
research topics, venues, and top cited papers,” Computer Science Review, vol. 27. pp. 16–32, 2018,
doi: 10.1016/j.cosrev.2017.10.002.
[11] D. M. E. D. M. Hussein, “A survey on sentiment analysis challenges,” J. King Saud Univ. - Eng.
Sci., vol. 30, no. 4, pp. 330–338, 2018, doi: 10.1016/j.jksues.2016.04.002.
[12] U. I. Larasati, M. A. Muslim, R. Arifudin, and A. Alamsyah, “Improve the Accuracy of Support
Vector Machine Using Chi Square Statistic and Term Frequency Inverse Document Frequency on
Movie Review Sentiment Analysis,” Sci. J. Informatics, vol. 6, no. 1, pp. 138–149, May 2019, doi:
10.15294/sji.v6i1.14244.
[13] Y. Gao, M. Gong, Y. Xie, and A. K. Qin, “An Attention-Based Unsupervised Adversarial Model
for Movie Review Spam Detection,” IEEE Trans. Multimed., vol. 23, pp. 784–796, 2021, doi:
10.1109/TMM.2020.2990085.
[14] S. Fransiska, R. Rianto, and A. I. Gufroni, “Sentiment Analysis Provider By.U on Google Play
Store Reviews with TF-IDF and Support Vector Machine (SVM) Method,” Sci. J. Informatics, vol.
7, no. 2, pp. 203–212, Nov. 2020, doi: 10.15294/SJI.V7I2.25596.
[15] N. P. Ririanti and A. Purwinarko, “Implementation of Support Vector Machine Algorithm with
Correlation-Based Feature Selection and Term Frequency Inverse Document Frequency for
Sentiment Analysis Review Hotel,” Sci. J. Informatics, vol. 8, no. 2, pp. 297–303, Nov. 2021, doi:
10.15294/sji.v8i2.29992.
[16] D. Apriliani, T. Abidin, E. Sutanta, A. Hamzah, and O. Somantri, “SentiHotel: a sentiment analysis
application of hotel services using an optimized neural network,” Bull. Electr. Eng. Informatics,
vol. 10, no. 3, Jun. 2021, doi: 10.11591/eei.v10i3.3040.
[17] O. Somantri and D. Apriliani, “Opinion mining on culinary food customer satisfaction using naïve
bayes based-on hybrid feature selection,” Indones. J. Electr. Eng. Comput. Sci., vol. 15, no. 1, pp.
468–475, Jul. 2019, doi: 10.11591/ijeecs.v15.i1.pp468-475.
[18] H. T. Ismet, T. Mustaqim, and D. Purwitasari, “Aspect Based Sentiment Analysis of Product
Review Using Memory Network,” Sci. J. Informatics, vol. 9, no. 1, pp. 73–83, May 2022, doi:
10.15294/sji.v9i1.34094.
[19] Y. Wang and X. Li, “Mining Product Reviews for Needs-Based Product Configurator Design: A
Transfer Learning-Based Approach,” IEEE Trans. Ind. Informatics, vol. 17, no. 9, pp. 6192–6199,
Sep. 2021, doi: 10.1109/TII.2020.3043315.
[20] Z. Zhao, J. Wang, H. Sun, Y. Liu, Z. Fan, and F. Xuan, “What Factors Influence Online Product

Scientific Journal of Informatics, Vol. 10, No.2, Aug 2023 | 237

 Sales? Online Reviews, Review System Curation, Online Promotional Marketing and Seller
Guarantees Analysis,” IEEE Access, vol. 8, pp. 3920–3931, 2020, doi:
10.1109/ACCESS.2019.2963047.
[21] R. K. Behera, M. Jena, S. K. Rath, and S. Misra, “Co-LSTM: Convolutional LSTM model for
sentiment analysis in social big data,” Inf. Process. Manag., vol. 58, no. 1, p. 102435, Jan. 2021,
doi: 10.1016/j.ipm.2020.102435.
[22] W. Liao, B. Zeng, J. Liu, P. Wei, X. Cheng, and W. Zhang, “Multi-level graph neural network for
text sentiment analysis,” Comput. Electr. Eng., vol. 92, p. 107096, Jun. 2021, doi:
10.1016/j.compeleceng.2021.107096.
[23] G. A. Ruz, P. A. Henríquez, and A. Mascareño, “Sentiment analysis of Twitter data during critical
events through Bayesian networks classifiers,” Futur. Gener. Comput. Syst., vol. 106, pp. 92–104,
May 2020, doi: 10.1016/j.future.2020.01.005.
[24] R. Blanquero, E. Carrizosa, P. Ramírez-Cobo, and M. R. Sillero-Denamiel, “Variable selection for
Naïve Bayes classification,” Comput. Oper. Res., vol. 135, p. 105456, Nov. 2021, doi:
10.1016/j.cor.2021.105456.
[25] Y. Yan, J. Chen, and Z. Wang, “Mining public sentiments and perspectives from geotagged social
media data for appraising the post-earthquake recovery of tourism destinations,” Appl. Geogr., vol.
123, p. 102306, Oct. 2020, doi: 10.1016/j.apgeog.2020.102306.
[26] E. Park, J. Park, and M. Hu, “Tourism demand forecasting with online news data mining,” Ann.
Tour. Res., vol. 90, p. 103273, Sep. 2021, doi: 10.1016/j.annals.2021.103273.
[27] T. Ali, B. Marc, B. Omar, K. Soulaimane, and S. Larbi, “Exploring destination’s negative e-
reputation using aspect based sentiment analysis approach: Case of Marrakech destination on
TripAdvisor,” Tour. Manag. Perspect., vol. 40, p. 100892, Oct. 2021, doi:
10.1016/j.tmp.2021.100892.
[28] B. Sohrabi, I. Raeesi Vanani, N. Nasiri, and A. Ghasemi Rud, “A predictive model of tourist
destinations based on tourists’ comments and interests using text analytics,” Tour. Manag.
Perspect., vol. 35, p. 100710, Jul. 2020, doi: 10.1016/j.tmp.2020.100710.
[29] S. Sun, C. Luo, and J. Chen, “A review of natural language processing techniques for opinion
mining systems,” Inf. Fusion, vol. 36, pp. 10–25, 2017, doi: 10.1016/j.inffus.2016.10.004.
[30] A. K. Uysal and S. Gunal, “The impact of preprocessing on text classification,” Inf. Process.
Manag., vol. 50, no. 1, pp. 104–112, Jan. 2014, doi: 10.1016/j.ipm.2013.08.006.
[31] J. Camacho and A. Ferrer, “Cross-validation in PCA models with the element-wise k-fold (ekf)
algorithm: Practical aspects,” Chemom. Intell. Lab. Syst., vol. 131, pp. 37–50, 2014, doi:
10.1016/j.chemolab.2013.12.003.
[32] P. Cichosz, “Naïve Bayes classifier,” in Data Mining Algorithms, Chichester, UK: John Wiley &
Sons, Ltd, 2015, pp. 118–133.
[33] S. Sona. D, S. Asha D, and B. Samarjeet, “5 Assimilate Machine Learning Algorithms in Big Data
Analytics: Review,” in Applications of Machine Learning in Big-Data Analytics and Cloud
Computing, River Publishers, 2021, pp. 81–114.
[34] S. Ruan, H. Li, C. Li, and K. Song, “Class-specific deep feature weighting for naïve bayes text
classifiers,” IEEE Access, vol. 8, pp. 20151–20159, 2020, doi: 10.1109/ACCESS.2020.2968984.
[35] A. Luque, M. Mazzoleni, A. Carrasco, and A. Ferramosca, “Visualizing Classification Results:
Confusion Star and Confusion Gear,” IEEE Access, vol. 10, pp. 1659–1677, 2022, doi:
10.1109/ACCESS.2021.3137630.
[36] M. Heydarian, T. E. Doyle, and R. Samavi, “MLCM: Multi-Label Confusion Matrix,” IEEE
Access, vol. 10, pp. 19083–19095, 2022, doi: 10.1109/ACCESS.2022.3151048.
[37] T. Reineking, “Active classification using belief functions and information gain maximization,”
Int. J. Approx. Reason., vol. 72, pp. 43–54, May 2016, doi: 10.1016/j.ijar.2015.12.005.
[38] Y. Gao, Y. Feng, and J. Tan, “Exploratory study on cognitive information gain modeling and
optimization of personalized recommendations for knowledge reuse,” J. Manuf. Syst., vol. 43, pp.
400–408, Apr. 2017, doi: 10.1016/j.jmsy.2017.01.003.
[39] D. Apriliani, T. Abidin, E. Sutanta, A. Hamzah, and O. Somantri, “Sentiment analysis for
assessment of hotel services review using feature selection approach based-on decision tree,” Int.
J. Adv. Comput. Sci. Appl., vol. 11, no. 4, pp. 240–245, Jun. 2020, doi:
10.14569/IJACSA.2020.0110432.
[40] F. Gorunescu, Data Mining: Concepts, Models and Techniques, vol. 12. Berlin, Heidelberg:
Springer Berlin Heidelberg, 2011.

238 | Scientific Journal of Informatics, Vol. 10, No. 3, Aug 2023