I. J.

Education and Management Engineering, 2023, 1, 29-34

Published Online on February 8, 2023 by MECS Press (http://www.mecs-press.org/)
DOI: 10.5815/ijeme.2023.01.04

Intelligent Model for Smartphone Addiction

Assessment in University Students using Android
Application and Smartphone Addiction Scale
Anshika Arora*
Department of Computer Science and Engineering, Netaji Subhas University of Technology, New Delhi-110078, India
Email: [email protected]
ORCID iD: https://orcid.org/0000-0003-0414-1874
*Corresponding author

Pinaki Chakraborty
Department of Computer Science and Engineering, Netaji Subhas University of Technology, New Delhi-110078, India
Email: [email protected]
ORCID iD: https://orcid.org/0000-0002-2010-8022

M.P.S. Bhatia
Department of Computer Science and Engineering, Netaji Subhas University of Technology, New Delhi-110078, India
Email: [email protected]
ORCID iD: https://orcid.org/0000-0001-7190-9770

Aditya Puri
Department of Electronics and Communication Engineering, Netaji Subhas University of Technology, New Delhi-
110078, India
Email: [email protected]

Received: 17 July, 2022; Revised: 06 September, 2022; Accepted: 16 October, 2022; Published: 08 February, 2023

Abstract: Smartphones have been owned and used ubiquitously in all facets of society utilized for a wide number of
tasks such as calling and messaging, social media, surfing as well as for entertainment. Spending a large amount of time
on smartphone might lead to a dependence on it for a variety of purposes. This study uses objective measures of real
time smartphone usage features to assess smartphone addiction. A purpose built android application to collect real time
smartphone usage has been developed and linear classification models namely Support Vector Machine and Logistic
Regression are used to predict smartphone addiction among university students. Furthermore, correlation and
information gain measures are used to identify most vital features of smartphone usage which contribute maximum in
assessment of smartphone addiction. It has been observed that both the linear models give worthy performance with
more than 80% of accuracy. Also, the most important technical features impacting smartphone addiction are longest
session spent for entertainment, total time used for communication, longest session spent for communication, longest
session spent for work, total time used for entertainment, longest session for news and surfing, and data usage in other
activities.

Index Terms: Smartphone addiction, machine learning, linear classification, android application.

1. Introduction

Smartphone usage has become inevitable with phones becoming an integral part of lifestyle serving multiple
purposes such as communications, social networking, media streaming, information retrieval, etc. Spending significant
amounts of time using smartphones for numerous activities potentially develop dependency on them for numerous
purposes which may lead to addictive usage and development of associated psychological disorders [1,2]. There are
many impairments associated with excessive and addictive used of smartphones such as increase in the risk of sleeping
disturbance and stress, sensation of being less physically active, fall in mental health and even depression [3]. Novel

This work is open access and licensed under the Creative Commons CC BY License. Volume 13 (2023), Issue 1
 Intelligent Model for Smartphone Addiction Assessment in University Students using Android Application and Smartphone
Addiction Scale

psychological variables such as nomophobia and Fear of Missing Out (FOMO) arising from smartphone addiction have
also been studied [2,4]. Researchers suggest that smartphone use can be detrimental if it forms an addiction via abuse of
technology.
Smartphone addiction is conventionally assessed via smartphone addiction scales, most of which rely on self-
reported data obtained through questionnaires [5]. However, a research restriction is that self-report based smartphone
addiction scales cannot reliably measure the actual phone usage. [5]. Moreover, this pairs with other shortcomings of
these type of scales such as unconscientious or missing responses, and wrong interpretations of questions which lead to
inaccurate information. Furthermore, filling of questionnaires is a cumbersome task for participants due to a large
number of questions. Purpose-built smartphone applications that can track phone usage attributes in order to provide
accurate measures of real-time usage patterns can provide an intervention for this. Subjective data from questionnaires
can be replaced by application extracted objective measures of smartphone usage features.
Motivated by this, the objective of this study is to use real time smartphone usage data to predict smartphone
addiction using supervised machine learning. A purpose built android application namely “UsageStats” has been
developed with the capability to track real time smartphone usage patterns among multiple attributes such as duration of
smartphone use and internet data used over the past 30 days. The applications installed in the smartphone are
categorized into nine types namely Social Media, Communication, Entertainment, Productivity, News & Surfing,
Gaming, Work, Photos & Camera, and Others. Patterns of usage of various categories of applications are attributed in
features such as category total time used, category sessions, category longest session, and data used in that category.
Smartphone addiction is labeled using the measures of most commonly used smartphone addiction scale, i.e.
Smartphone Addiction Scale-Short Version (SAS-SV) designed by Known et al. [6]. Linear classifiers including
Support Vector Machine (SVM) and Logistic Regression (LR) are used to predict smartphone addiction. Also, the most
contributing attributes of smartphone usage are identified using correlation attribute evaluator and Information Gain (IG)
attribute evaluator using predefined thresholds.

2. Related Work

The conventional approach for assessment of smartphone addiction is based on questionnaire based smartphone
addiction scales which have been discussed in existing studies [2,5]. Soft computing techniques have been applied to
these self-reported data by a number of researchers to provide a technically sophisticated diagnosis of smartphone
addiction. Shin et al. [7] used a variety of mobile phone usage data to discover a number of variables in order to build
machine learning methods for automated prediction of problematic smartphone use, such as Nave Bayes, Support
Vector Machine, and AdaBoost. Supervised machine learning models based on K-Nearest Neighbour, Decision Tree,
Nave Bayes, and Support Vector Machine were proposed for smartphone addiction evaluation [8]. However the
accuracy and reliability of these results is subject to biasness in information filled.
Recent studies have suggested successful interventions of real time smartphone usage tracking for analysis of
behavioral patterns and assessment of problematic use [1,9]. Soft computing methodologies are being used on these
real-time smartphone extracted data to provide an accurate diagnosis of smartphone addiction. Ellis et al. [10]
implemented an unsupervised machine learning algorithm, i.e. K-Means clustering to group the users with similar
smartphone usage patterns. They used the usage behavior pattern retrieved from Apple’s Screen Time application which
logs a series of behavioral screen time metrics. Other psychological variables arising from smartphone addiction have
also been assessed by researchers using real time smartphone extracted usage patterns and soft computing techniques.
Arora et al. [2] utilized Gaussian mixture clustering algorithm to identify nomophobic behavior in users with similar
smartphone usage pattern retrieved via smartphone application. In another study, Arora et al. [11] extracted real time
smartphone data via an android application for prediction of nomophobia severity using supervised machine learning
algorithms such as Random Forest, Decision Tree, Support Vector Machines, Naïve Bayes And K-Nearest Neighbor.
Elhai et al. [12] implemented supervised machine learning algorithms in order to detect problematic smartphone use
severity among undergraduate students in China.

3. Materials and Methods

The materials and methods used in this study are described in the following subsections.
3.1 Android Application: UsageStats
In order to collect real time smartphone usage data of participants, an android application namely “UsageStats” has
been developed which automatically collects data from the smartphones of user. Fig. 1 presents the UsageStats screen.

30 Volume 13 (2023), Issue 1

 Intelligent Model for Smartphone Addiction Assessment in University Students using Android Application and Smartphone
Addiction Scale

Fig. 1. Smartphone usage during the month

The application tracks overall smartphone usage as well as usage of various smartphone applications including the
in-built applications and third party applications. The features of smartphone usage directly collected by UsgeStats are
listed in Table 1.

Table 1. Features Extracted by Smartphone Application

S. No. Attribute Description
1. Total_time Total duration of smartphone use for the last 30 days
2. Total_data_used Sum of internet data used by all apps during the smartphone use in last 30 days
For each application
1. Time_used Total duration of the application usage for last 30 days
2. Times_open (sessions) Total number of sessions of the application in last 30 days
3. Longest_session Longest duration in last 30 days for which the application is used in one launch
4. Data_used Sum of mobile and Wi-Fi internet data that has been used during the application
usage in last 30 days
5. Tag Category to which the application belongs

The application is sent to undergraduate students of Netaji Subhas University of Technology and the records are
maintained in an online database.
3.1.1 Smartphone Usage Attribute Set
After data collection using the UsageStats, the applications are categorized into nine tags viz. Social Meida,
Communication, Entertainment, Productivity, News & Surfing, Gaming, Work, Photos & Camera, and Others. These
tags are created by sorting and modifying the default tags extracted by the developed application. Applications are
categorized as follows:

1. Social Media includes applications such as Twitter, Tumblr, Facebook, Instagram, Reddit.
2. Communication includes messaging applications and call applications like Snapchat, WhatsApp, Telegram,
Truecaller, messenger, Discord and system applications such as contacts, phone, messaging, contacts and
dialer etc.
3. Entertainment combines system applications such as movies & videos and music & audio as well as the third
party video and music applications like Youtube, Netflix, Sonyliv, Prime video, Hotstar, Spotify, FM Radio
Gaana, etc.
4. Productivity includes the applications like Paytm, Gpay, Docs, Sheets, etc.
5. News & Surfing includes various browsers such as Google, Chrome, Browser, Mi Browser, Internet and
others as well as news and magazine applications like Inshorts, Quora, Google news, etc.
6. Gaming tag includes third party gaming applications like Battleground.

Volume 13 (2023), Issue 1 31

 Intelligent Model for Smartphone Addiction Assessment in University Students using Android Application and Smartphone
Addiction Scale

7. Work tag is maintained for all the work and study related applications including Gmail, Mail, LinkedIn,
Classroom, Zoom, Unacademy, Udemy Meet, etc.
8. Photos & Camera include system applications like Albums, Camera, Photos, Gallery and third party
applications like Picsart.
9. Others is the tag for applications belonging to none of the abovementioned categories. It includes applications
like Files, numerous shopping applications, drive and maps applications and various food delivery
applications.

After categorizing and sorting the applications into their respective tags, a dataset is created which for each user
contains attributes of overall smartphone usage such as total_time in hours and total_data_used in GB and attributes for
every tag including time_used in minutes, times_open (sessions), longest_session in minutes, and data_used in MB.
3.1.2 Normalization
The smartphone usage attribute set is normalized using the Min-Max normalization. For each attribute it
transforms its values to a decimal between 0 and 1 which works as given in equation 1.

𝑥𝑎 −𝑚𝑖𝑛𝑎
𝑥 ′𝑎 = (1)
𝑚𝑎𝑥𝑎 −𝑚𝑖𝑛𝑎

where, 𝑥𝑎 is a instance value belonging to attribute a,

𝑚𝑖𝑛𝑎 is the minimum value over attribute a
𝑚𝑎𝑥𝑎 is the maximum value over attribute a.
3.2 Labeling using SAS-SV
SAS-SV designed by by Known et al. [6] is the most commonly used scale for assessment of smartphone addiction
[13]. This study uses SAS-SV to collect responses from students about their perception on their dependence on
smartphones. SAS-SV contains 10 questions each of which has to be answered on a 6-point Likert scale. The sum of the
answers is the indicator of presence of smartphone addiction with a cut off value of 33 for girls and 31 for boys. This is
used for labeling students with smartphone addiction.
3.3 Classification
The dataset containing real time smartphone usage features with labels derived from SAS-SV is fed as input to two
linear classification models namely, SVM and Logistic Regression. These models are chosen as they are the most
suitable model for small datasets [14]. The performances of classification models are evaluated with accuracy and F1-
score.
3.4 Identification of Most Contributing Smartphone Usage Attributes
This study uses statistical attribute evaluator methods which evaluate the relationship between each input variable
and the target variable in order to use scoring measures as a basis to filter those input variables which contribute
maximum for classification [15]. The methods used are Correlation attribute evaluator and Information Gain attribute
evaluator. Correlation attribute evaluator evaluates the worth of an attribute by measuring the Pearson's correlation
between it and the class [16,17]. IG attribute evaluator evaluates the worth of an attribute by measuring the information
gain with respect to the class [16,18]. For each attribute evaluator, top 5 smartphone usage features are identified which
contribute the most in assessment of smartphone addiction.

4. Results

The dataset is labeled using SAS-SV and evaluated using the real time attributes extracted via android application.
Linear classification performance is evaluated using 5-fold cross validation.

Table 2. Classification Performance

Accuracy F1 Score
Linear SVM 81.33 0.810
Logistic Regression 82.67 0.827

The most important smartphone usage attributes contributing maximum in assessment of smartphone addiction are
identified using correlation and information gain. Table 3 and Table 4 present the top five attributes extracted using the
correlation evaluator and the information gain evaluator respectively.

32 Volume 13 (2023), Issue 1

 Intelligent Model for Smartphone Addiction Assessment in University Students using Android Application and Smartphone
Addiction Scale

Table 3. Top 5 Attributes Identified Using Correlation Evaluator

S. No. Attributes Correlation Score
1. Entertainment Longest Session 0.4129
2. Communication Total Time 0.3987
3. Communication Longest Session 0.3496
4. Work Longest Session 0.348
5. Entertainment Total Time 0.3475

Table 4. Top 5 Attributes Identified Using Information Gain Evaluator

S. No. Attributes IG Score
1. News and Surfing Longest Session 0.4129
2. Communication Total Time 0.3987
3. Communication Longest Session 0.3496
4. Entertainment Longest Session 0.348
5. Others Data Usage 0.3475

5. Conclusion

This study proposes a model for assessment of smartphone addiction in university students using the objective
measures of real time smartphone usage and supervised machine learning. An android application is developed to
collect real time smartphone usage features and linear classification models namely Support Vector Machine and
Logistic regression are used to predict smartphone addiction among university students. Smartphone Addiction Scale-
Short Version is used to label smartphone addiction among university students. Support Vector Machine predicts
smartphone addiction with a superlative accuracy of 81.33% while Logistic Regression gives accuracy of 82.67%. Most
contributing features of real time smartphone usage for assessment of smartphone addiction have also been identified
using correlation and information gain measures.
The easy availability of smartphones and the Internet and the flexible schedules of millennials including students
are leading them to have addictive smartphone usage behaviors [11]. Online education which involves use of digital
technologies including smartphones has become useful, yet stressful practice for university students after the COVID-
19 pandemic. However, overuse of smartphones and other digital technologies are known to cause fatigue [19] and
affect academic performance of students [20]. Therefore, universities should make judicious use of these technologies.

References

[1] Tossell, C., Kortum, P., Shepard, C., Rahmati, A. and Zhong, L., 2015. Exploring smartphone addiction: insights from long-
term telemetric behavioral measures. International Journal of Interactive Mobile Technologies, 9(2).
[2] Arora, A., Chakraborty, P. and Bhatia, M.P.S., 2021. Problematic use of digital technologies and its impact on mental health
during COVID-19 pandemic: assessment using machine learning. In Emerging Technologies During the Era of COVID-19
Pandemic (pp. 197-221). Springer, Cham.
[3] Vernon, L., Modecki, K.L. and Barber, B.L., 2018. Mobile phones in the bedroom: Trajectories of sleep habits and subsequent
adolescent psychosocial development. Child development, 89(1), pp.66-77.
[4] Arora, A. and Chakraborty, P., 2020. Diagnosis, prevalence and effects of nomophobia-A review. Psychiatry Research, 288,
pp.112975-112975.
[5] Harris, B., Regan, T., Schueler, J. and Fields, S.A., 2020. Problematic mobile phone and smartphone use scales: A systematic
review. Frontiers in Psychology, 11, p.672.
[6] Kwon, M., Kim, D.H., Cho, H. and Yang, S., 2013. SAS-SV (also known as: Smartphone Addiction Scale-Short
Version)(appears in: The Smartphone Addiction Scale: Development and Validation of a Short Version for Adolescents.)
Copyright: Creative Commons License.
[7] Shin, C. and Dey, A.K., 2013, September. Automatically detecting problematic use of smartphones. In Proceedings of the 2013
ACM international joint conference on Pervasive and ubiquitous computing (pp. 335-344).
[8] Lawanont, W. and Inoue, M., 2017, June. A development of classification model for smartphone addiction recognition system
based on smartphone usage data. In International Conference on Intelligent Decision Technologies (pp. 3-12). Springer, Cham.
[9] Arora, A., Chakraborty, P. and Bhatia, M.P.S., 2021. Intervention of Wearables and Smartphones in Real Time Monitoring of
Sleep and Behavioral Health: An Assessment Using Adaptive Neuro-Fuzzy Technique. Arabian Journal for Science and
Engineering, pp.1-26.
[10] Ellis, D.A., Davidson, B.I., Shaw, H. and Geyer, K., 2019. Do smartphone usage scales predict behavior?. International
Journal of Human-Computer Studies, 130, pp.86-92.
[11] Arora, A., Chakraborty, P. and Bhatia, M.P.S., 2021. Real Time Smartphone Data for Prediction of Nomophobia Severity using
Supervised Machine Learning. AIJR Proceedings, pp.78-83.
[12] Elhai, J.D., Yang, H., Rozgonjuk, D. and Montag, C., 2020. Using machine learning to model problematic smartphone use
severity: The significant role of fear of missing out. Addictive behaviors, 103, p.106261.
[13] Harris, B., McCredie, M. and Fields, S., 2020. Examining the psychometric properties of the smartphone addiction scale and its
short version for use with emerging adults in the US. Computers in Human Behavior Reports, 1, p.100011.

Volume 13 (2023), Issue 1 33

 Intelligent Model for Smartphone Addiction Assessment in University Students using Android Application and Smartphone
Addiction Scale

[14] Dreiseitl, S. and Ohno-Machado, L., 2002. Logistic regression and artificial neural network classification models: a
methodology review. Journal of biomedical informatics, 35(5-6), pp.352-359.
[15] Kumar, A. and Arora, A., 2019, February. A filter-wrapper based feature selection for optimized website quality prediction.
In 2019 Amity International Conference on Artificial Intelligence (AICAI) (pp. 284-291). IEEE.
[16] Gnanambal, S., Thangaraj, M., Meenatchi, V.T. and Gayathri, V., 2018. Classification algorithms with attribute selection: an
evaluation study using WEKA. International Journal of Advanced Networking and Applications, 9(6), pp.3640-3644.
[17] Sugianela, Y. and Ahmad, T., 2020, February. Pearson correlation attribute evaluation-based feature selection for intrusion
detection system. In 2020 International Conference on Smart Technology and Applications (ICoSTA) (pp. 1-5). IEEE.
[18] Win, T.Z. and Kham, N.S.M., 2019, February. Information gain measured feature selection to reduce high dimensional data.
In Seventeenth International Conference on Computer Applications (ICCA 2019) (Vol. 68, No. 73, pp. 1-5).
[19] Shin, J. and Shin, M., 2016. To be connected or not to be connected? Mobile messenger overload, fatigue, and mobile
shunning. Cyberpsychology, Behavior, and Social Networking, 19(10), pp.579-586.
[20] Liu, X., Luo, Y., Liu, Z. Z., Yang, Y., Liu, J., & Jia, C. X. (2020). Prolonged mobile phone use is associated with poor
academic performance in adolescents. Cyberpsychology, Behavior, and Social Networking, 23(5), 303-311.

Authors’ Profiles

Anshika Arora received her B.Tech. in computer science from the University of Delhi and M.Tech. in software
engineering (gold medallist) from Delhi Technological University. She is currently pursuing Ph.D. at Netaji
Subhas University of Technology, New Delhi. Her research interests include behavioral health analysis, web
analytics using soft computing techniques.

Pinaki Chakraborty is an assistant professor at the Department of Computer Science and Engineering, Netaji
Subhas University of Technology, New Delhi. He received his B.Tech. from Indraprastha University, and his
M.Tech. and Ph.D. from Jawaharlal Nehru University. He has published more than 90 research papers in journals
and conference of repute. His area or research includes systems software and educational and societal use of
software.

M.P.S. Bhatia is professor in Department of Computer Science and Engineering, Netaji Subhas University of
Technology. He is Ph.D. in computer engineering with more than twenty years of research experience. His area of
research includes data mining, machine learning, cyber security, social network analysis, computer vision, smart
health and big data analytics.

Aditya Puri is pursuing B.E. in electronics and communication engineering at Netaji Subhas University of
Technology. His interests lie in android development and Java backend development.

How to cite this paper: Anshika Arora, Pinaki Chakraborty, M.P.S. Bhatia, Aditya Puri, "Intelligent Model for Smartphone
Addiction Assessment in University Students using Android Application and Smartphone Addiction Scale", International Journal of
Education and Management Engineering (IJEME), Vol.13, No.1, pp. 29-34, 2023. DOI:10.5815/ijeme.2023.01.04

34 Volume 13 (2023), Issue 1