Predicting ratings of social media feeds: combining latent-factors and emotional aspects for improving performance of different classifiers
| DOI | https://doi.org/10.1108/AJIM-12-2021-0357 |
| Published date | 10 May 2022 |
| Date | 10 May 2022 |
| Pages | 1126-1150 |
| Subject Matter | Library & information science,Information behaviour & retrieval,Information & knowledge management,Information management & governance,Information management |
| Author | Arghya Ray,Pradip Kumar Bala,Nripendra P. Rana,Yogesh K. Dwivedi |
Predicting ratings of social media
feeds: combining latent-factors
and emotional aspects for
improving performance of
different classifiers
Arghya Ray
Management Information Systems and Analytics,
International Management Institute Kolkata, Kolkata, India
Pradip Kumar Bala
Department of Information Systems and Business Analytics,
Indian Institute of Management Ranchi, Ranchi, India
Nripendra P. Rana
College of Business and Economics, Qatar University, Doha, Qatar, and
Yogesh K. Dwivedi
Swansea University, Swansea, UK and
Department of Management, Symbiosis Institute of Business Management,
Pune and Symbiosis International (Deemed University), Pune, India
Abstract
Purpose –The widespread acceptance of various social platforms has increased the number of users posting
about various services based on their experiences about the services. Finding out the intended ratings of social
media (SM) posts is important for both organizations and prospective users since these posts can help in
capturing the user’s perspectives. However, unlike merchant websites, the SM posts related to the service-
experience cannot be rated unless explicitly mentioned in the comments. Additionally, predicting ratings can
also help to build a database using recent comments for testing recommender algorithms in various scenarios.
Design/methodology/approach –In this study, the authors have predicted the ratings of SM posts using
linear (Naı€veBayes, max-entropy) and non-linear (k-nearest neighbor, k-NN) classifiers utilizing combinations
of different features, sentiment scores and emotion scores.
Findings–Overall, the results of this study reveal that the non-linear classifier (k-NN classifier) performed
better than the linear classifiers (Naı€ve Bayes, Max-entropy classifier). Results also show an improvement
of performance where the classifier was combined with sentiment and emotion scores. Introduction of the
feature “factors of importance”or “the latent factors”also show an improvement of the classifier
performance.
Originality/value –This study provides a new avenue of predicting ratings of SM feeds by the use of
machine learning algorithms along with a combination of different features like emotional aspects and latent
factors.
Keywords Emotional aspects, K-nearest neighbors, Max-entropy, Naı€vebayes, Rating prediction,
Social media feeds
Paper type Research paper
AJIM
74,6
1126
The infrastructural support provided by IMI Kolkata, IIM Ranchi, Swansea University, Symbiosis
International (Deemed University) and Qatar University in completing this paper is gratefully
acknowledged.
The current issue and full text archive of this journal is available on Emerald Insight at:
https://www.emerald.com/insight/2050-3806.htm
Received 2 December 2021
Revised 3 January 2022
30 March 2022
Accepted 9 April 2022
Aslib Journal of Information
Management
Vol. 74 No. 6, 2022
pp. 1126-1150
© Emerald Publishing Limited
2050-3806
DOI 10.1108/AJIM-12-2021-0357
1. Introduction
The availability of Internet and the widespread acceptance of social media (SM) platforms
(e.g. Facebook, Twitter, etc.) has led to the growth of online service providers and an increase
in overall competition in almost every sector (Ray et al., 2019a). Knowing what a customer
feels about a product or service based on the reviews or ratings provided by customers is of
great strategic importance for organizations (Black and Kelley, 2009). While online reviews
can impact sales (Li et al., 2020), online ratings reflect quality and consumer satisfaction
(Engler et al., 2015) associated with a product. Both online reviews and ratings influence the
decisions of prospective customers (Cui et al., 2012;Kostyra et al., 2016;Chatterjee, 2019;Ray
et al., 2020b). Apart from posting reviews on merchandise or organizational pages, customers
also post their views about a product or service in SM due to the increasing popularity of SM
platforms and the easy means of sharing information (Simon et al., 2015;Chaffey, 2019).
Customer behavior can however be affected by several factors. Latent factor refers to the
unobserved factor/s which although not observable can affect customer behavior. These
latent factors are however important for deciding customer behavior, and we have termed it
as “factors-of-importance”for proposing a new feature for classifiers. Comments posted on
SM about various products/services are rich in information (Chatterjee, 2019). A proper
analysis of SM posts about products/services can reveal factors (latent factors) affecting
customer behavior (Ray et al., 2019a,b) and can also throw light on the underlying sentiments
and emotions involved (Chatterjee, 2019). Although researchers (e.g. Mukherjee and Bala,
2017) have worked on features such as content words and function words, it increases the
time and space complexity because of the use of a large bag-of-words for constant comparison
and testing. However, no studies have utilized the “factors-of-importance”as a feature.
Although researchers (e.g. Dooms et al., 2014;Prasetyo and Winarko, 2016) have worked on
rating prediction, the studies lack good prediction accuracy (Ray et al., 2020a). The need for
improving the accuracy of rating-prediction and to devise a way to reduce space and time
complexity of classifiers motivated us to work on these gaps.
The business problems that drive this research are as follows. First, there is a lack of
proper research on accurate rating prediction from SM posts about various products/
services. For effective SM marketing, organizations need to understand consumer
perspectives better and engage with them on SM platforms (Mosley, 2018). Due to the
large number of SM posts about products/services on SM platforms (Simon et al., 2015), it is
not easy for service providers to segment the negative SM posts from the good ones (Oheix,
2018). But the SM posts have both strategic and marketing importance. When a new
prospective customer reads a comment or post about a certain product/service, their
decisions can be affected. Additionally, it has also been found that people do not usually
spend much time going through a post (Patel, 2016;Read, 2016) but rather take a glance
through the catchy words. However, when the ratings are available, it becomes easier for
users to take a decision. Unlike merchandise websites, ratings are not available for SM posts.
Thus, SM posts regarding a product/service will not be of much value to users who do not
love reading lengthy posts. However, for the providers, each and every post is important for
understanding customers’views and for solving the problems that users face. Inability to
address the issues instantly may lead to loss of customers. Detecting ratings thus helps to
provide a view of the users’feelings and hence becomes an important part of customer
service. Second, deriving ratings from SM posts will help in generating datasets from most
recent posts for training recommender systems (RSs) (Chambua and Niu, 2021). This will help
to reduce the dependence on older datasets like MovieLens (Dooms et al., 2013). Third, since
people post their views about a product/service in different platforms, detecting ratings and
combing them from different platforms will help users gain a better understanding about the
product/service. This will also help providers to understand how they are performing in the
market. Fourth, existing research has used features, like, content words, function words, etc.
Predicting
ratings of
social media
feeds
1127
To continue reading
Request your trial