A Multi‑Layer Perceptron Model for Predicting Smartphone

Addition Levels

S. Nasrin, M. Fahimunnisa, Aslam Shareef, S. Ananya Jyothi and P. Jasmin

Department of Computer Science, Ravindra College of Engineering for Women, Venkayapalle, Pasupala, NH 340C,

Nandikotkur Rd, Kurnool, Andhra Pradesh, India

Keywords: Machine Learning, Behavioral Analysis, Predictive Model, Digital Well‑Being, Classification Algorithms,

User Behavior, Artificial Intelligence, Neural Networks (ANNs), Mental Health.

Abstract: Smartphone addiction is a serious issue, with adverse effects on mental health, academic achievement, sleep

and social relationships. Conventional self-reported questionnaires tend to be inaccurate and susceptible to

bias, requiring automated approaches. This research suggests a machine learning model to forecast

smartphone addiction based on behavioral metrics like screen time, app usage, social media usage, call

duration, and phone unlock frequency. Psychological variables from well-validated questionnaires are also

included to enhance prediction accuracy. Different machine learning algorithms such as Decision Trees,

SVM, Random Forests, and Neural Networks are experimented with on a labeled dataset. Accuracy, precision,

recall, and F1-score are used to evaluate the models, and the results indicate that ensemble methods such as

Random Forests work best. The system allows real-time tracking of smartphone addiction risks. It provides

an early intervention proactive approach and management of addiction. It can be implemented into digital

health applications for users, educators, and healthcare providers. It ultimately seeks to enable healthier

smartphone use and digital well-being.

1 INTRODUCTION

In recent years, smartphones have transformed from

mere communication tools into indispensable digital

companions. Their integration into everyday life

ranging from social networking and entertainment to

education and business has created a deep reliance on

mobile devices. While smartphones offer immense

convenience and connectivity, their overuse has given

rise to a modern behavioral disorder known as

smartphone addiction. This addiction is characterized

by compulsive and excessive use of smartphones,

leading to negative impacts on mental health,

interpersonal relationships, and overall productivity.

Numerous studies have shown a significant

correlation between excessive smartphone usage and

issues such as anxiety, depression, sleep disturbances,

and poor academic or professional performance.

Particularly among adolescents and young adults, the

increasing screen time and engagement in digital

platforms have raised concerns about behavioral

dependence. Traditional diagnostic methods such as

surveys and psychological evaluations are effective

but often limited by subjective interpretation, delayed

analysis, and the inability to monitor behavioral

patterns in real-time.

The rise of artificial intelligence and machine

learning technologies presents an opportunity to

address this issue in a more dynamic and data-driven

manner. Machine learning algorithms can analyze

large volumes of behavioral data to uncover patterns

and detect signs of addictive behavior. By utilizing

smartphone usage metrics such as app usage duration,

screen time frequency, social media interactions, and

phone unlock counts along with psychological inputs,

machine learning models can provide an accurate and

automated assessment of smartphone addiction risk

levels.

This project aims to develop a machine learning

model capable of predicting smartphone addiction by

analyzing behavioral and psychological data. The

model is trained on datasets collected from

smartphone users, incorporating both usage logs and

responses from validated psychological

questionnaires. Various supervised learning

algorithms, including Decision Trees, Support Vector

Machines (SVM), Random Forest, and Neural

Networks, are employed and evaluated to determine

482

Nasrin, S., Fahimunnisa, M., Shareef, A., Jyothi, S. A. and Jasmin, P.

A Multi-Layer Perceptron Model for Predicting Smartphone Addition Levels.

DOI: 10.5220/0013915300004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 4, pages

482-486

ISBN: 978-989-758-777-1

the most effective method for classification and

prediction.

The goal of this study is not only to enhance the

early detection of smartphone addiction but also to

support the development of preventive tools and

digital well-being solutions. By integrating such a

system into mobile applications, users can receive

timely alerts and recommendations for healthier

smartphone habits. Additionally, educators,

psychologists, and health professionals can use the

insights generated by the model to guide individuals

toward more balanced digital lifestyles and reduce the

long-term consequences of technology

overdependence.

2 RELATED WORKS

Several studies have explored smartphone addiction

prediction using machine learning. Research by Kim

et al. (2018) used Decision Trees and Logistic

Regression on usage data, showing high correlation

with addiction levels. Wang et al. (2020) applied

SVM and Random Forest for addiction classification,

achieving considerable accuracy. Lee & Lee (2021)

improved predictions by combining behavioral and

psychological data, while Chen et al. (2022)

emphasized the role of feature selection in model

performance. This study builds upon existing

research by integrating behavioral metrics with

psychological assessments and evaluating multiple

machine learning algorithms for robust addiction

prediction.

3 METHODOLOGY

3.1 Theoretical Structure

The research methodology adopted for this study

involves a systematic approach to collecting,

processing, and analyzing smartphone usage and

psychological data to develop a machine learning

model capable of predicting smartphone addiction.

The first phase involves data collection from

participants using both primary and secondary

sources. Primary data is gathered through smartphone

usage monitoring applications that log features such

as screen time, app usage frequency, call and message

duration, and unlock counts. Simultaneously,

participants are asked to complete validated

psychological questionnaires such as the Smartphone

Addiction Scale (SAS), which helps in labeling the

addiction severity.

In the second phase, the collected data is

preprocessed to ensure consistency and reliability.

Data cleaning techniques are applied to remove

duplicates, handle missing values, and standardize

feature values. Feature engineering is performed to

select the most relevant attributes that contribute

significantly to addiction prediction. This phase also

includes normalization and encoding of categorical

data to prepare the dataset for machine learning

algorithms.

The third phase focuses on model development

and training. Several supervised learning algorithms

such as Decision Trees, Support Vector Machines

(SVM), Random Forest, Logistic Regression, and

Artificial Neural Networks are implemented using

programming tools like Python and machine learning

libraries such as Scikit-learn and TensorFlow. The

dataset is split into training and testing sets to evaluate

model performance. Hyperparameter tuning and

cross-validation techniques are applied to enhance the

accuracy and generalization of the models.

Finally, model evaluation is conducted using

performance metrics such as Accuracy, Precision,

Recall, F1-score, and ROC-AUC. The model with the

highest predictive performance is selected as the final

model. The results are interpreted to determine the

patterns and factors contributing most significantly to

smartphone addiction. The methodology ensures a

robust, data-driven, and repeatable process for

developing a predictive tool that can sbe. Figure 1

Shows the Schematic Flow of Theoretical Structure.

Figure 1: Schematic flow of theoretical structure.

A Multi-Layer Perceptron Model for Predicting Smartphone Addition Levels

483

3.2 Perceived Features

3.2.1 Data Collection and Processing

Data is collected through smartphone monitoring

apps and validated psychological scales such as the

Smartphone Addiction Scale (SAS). The dataset

undergoes preprocessing, including data cleaning,

feature selection, and encoding. Feature engineering

is applied to extract the most relevant attributes for

addiction prediction.

3.2.2 Machine Learning Model Development

Various supervised learning models are implemented,

including Decision Trees, SVM, Random Forest, and

Artificial Neural Networks. The dataset is split into

training and testing sets, with hyperparameter tuning

and cross-validation performed to enhance accuracy.

Model performance is evaluated using Accuracy,

Precision, Recall, F1-score, and ROC-AUC.

3.2.3 Statistical Analysis

Structural Equation Modeling (SEM) is used to assess

the relationship between smartphone usage behaviors

and addiction severity. Pearson’s correlation analysis

identifies the strongest predictors of addiction. The

final model is selected based on its predictive

accuracy and generalization capability.

4 PURPOSE

The proposed system aims to develop an intelligent,

data-driven, and automated model for the prediction

of smartphone addiction using machine learning

techniques. Unlike traditional systems that rely

heavily on self-reported data and static assessments,

this system leverages real-time smartphone usage

patterns and behavioral metrics to assess addiction

risk. By collecting and analyzing data such as screen

time, app usage frequency, unlock count, call

duration, and notifications, the system can detect

early signs of addiction without user intervention. A

core element of the proposed system is the integration

of machine learning algorithms to accurately classify

and predict smartphone addiction levels.

Additionally, the proposed system is designed to be

adaptive and personalized. It monitors behavioral

trends over time and adjusts risk prediction based on

a user’s changing habits. Users receive feedback in

the form of addiction risk levels (low, moderate, or

high) along with actionable suggestions such as

screen-time reduction tips, app usage control, and

wellness prompts. This feedback mechanism helps

users stay aware of their usage patterns and motivates

them to adopt healthier digital behaviors proactively.

Ultimately, the proposed system addresses the

limitations of existing solutions by providing an

intelligent, real-time, and scalable framework for

smartphone addiction prediction. It can be deployed

in mobile applications, institutional wellness

programs, or digital therapy platforms to support

individuals, students, and employees in maintaining

digital well-being. By identifying at-risk users early,

the system can contribute to preventive mental

healthcare and reduce the long-term impact of

smartphone addiction.

5 RESULT

The performance of the prediction model is evaluated

by comparing real and predicted mobile phone

addiction levels among students (Figure 2).

Furthermore, the effectiveness of different machine

learning techniques is assessed using R² values across

three model variants (Figure 3).

Figure 2: Comparison of real and predicted mobile phone

addiction levels among students.

The scatter plot presented in the image visualizes

the performance of a Multi-Layer Perceptron (MLP)

model in predicting smartphone addiction levels

among students. The X-axis represents the number of

students, while the Y-axis indicates their respective

smartphone addiction levels.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

484

Figure 3: R² value comparison of various machine learning

models across three model variants.

6 CONCLUSIONS

In today’s technology-driven world, smartphone

addiction has become a significant concern affecting

the mental health, productivity, and social well-being

of individuals, especially among youth. Traditional

methods of identifying addiction are limited by their

dependency on subjective responses and delayed

interventions. This project presents a modern solution

by utilizing machine learning techniques to automate

and enhance the accuracy of addiction prediction,

thus enabling timely awareness and preventive

actions.

The proposed system leverages real-time

smartphone usage data and psychological inputs to

build intelligent models capable of predicting

addiction levels effectively. By integrating various

supervised machine learning algorithms and

analyzing key behavioral features such as screen

time, app usage, and unlock frequency, the model

offers a more comprehensive and accurate assessment

of smartphone addiction risk.

Moreover, the system is designed to be adaptive,

scalable, and user-centric. It not only identifies high-

risk individuals but also provides personalized

suggestions and insights to help users regain control

over their smartphone usage. The use of hybrid

data—combining behavioral and psychological

parameters—enhances the depth of analysis, making

the system a powerful tool for digital wellness and

mental health awareness.

In conclusion, this machine learning-based

approach to smartphone addiction prediction marks a

significant step forward in addressing the growing

challenges of digital dependency. With further

development and real-world implementation, this

system has the potential to support individuals,

institutions, and healthcare professionals in

promoting healthier digital habits and improving

overall quality of life.

REFERENCES

Bianchi, A., & Phillips, J. G. (2005). Psychological

predictors of problem mobile phone use.

CyberPsychology & Behavior, 8(1), 39–51.

Chen, Y., Sun, X., & Zhao, J. (2022). Feature selection

methods for smartphone addiction prediction: A

comparative study. Journal of Computational Science,

58, 101512

Demirci, K., Akgönül, M., & Akpinar, A. (2015).

Relationship of smartphone use severity with sleep

quality, depression, and anxiety in university students.

Journal of Behavioral Addictions, 4(2), 85–92.

Jeong, S. H., Kim, H., Yum, J. Y., & Hwang, Y. (2016).

What type of content are smartphone users addicted to?

SNS vs. games. Computers in Human Behavior, 54,

10–17.

Kim, J. H., Seo, M., & David, P. (2015). Alleviating

depression only to become problematic mobile phone

users: Can face-to-face communication be the antidote?

Computers in Human Behavior, 51, 440-447.

Kwon, M., Kim, D. J., Cho, H., & Yang, S. (2013). The

Smartphone Addiction Scale: Development and

validation of a short version for adolescents. PLOS

ONE, 8(12), e83558.

Lee, Y. K., Chang, C. T., Lin, Y., & Cheng, Z. H. (2014).

The dark side of smartphone usage: Psychological

traits, compulsive behavior, and technostress.

Computers in Human Behavior, 31, 373–383.

Leung, L. (2008). Linking psychological attributes to

addiction and improper use of the mobile phone among

adolescents in Hong Kong. Journal of Children and

Media, 2(2), 93–113.

Lin, Y. H., Lin, Y. C., Lee, Y. H., Lin, P. H., Lin, S. H.,

Chang, L. R., & Kuo, T. B. (2015). Time distortion

associated with smartphone addiction: Identifying

smartphone addiction via a mobile application (App).

Journal of Psychiatric Research, 65, 139-145.

Liu, C., & Ma, J. L. (2020). Social media addiction and

burnout: The mediating roles of envy and social media

engagement. International Journal of Environmental

Research and Public Health, 17(10), 3723.

Roberts, J. A., & David, M. E. (2019). The social media

party: Fear of missing out (FoMO), social media

intensity, connection, and well-being. International

Journal of Human–Computer Interaction, 35(10), 909–

914.

Rozgonjuk, D., Saal, K., Täht, K., & Vassil, K. (2020).

Predicting problematic smartphone use with personality

traits and usage patterns. Heliyon, 6(11), e05448.

A Multi-Layer Perceptron Model for Predicting Smartphone Addition Levels

485

Samaha, M., & Hawi, N. S. (2016). Relationships among

smartphone addiction, stress, academic performance,

and satisfaction with life. Computers in Human

Behavior, 57, 321–325.

Suresh, A., & Ganapathi, V. (2021). Smartphone addiction

detection using data mining techniques: A survey.

International Journal of Scientific Research in

Computer Science, Engineering and Information

Technology, 7(1), 18–23.

Wang, P., Liu, T., & Zhang, J. (2020). Machine learning-

based detection of smartphone addiction using

behavioral data. International Journal of Environmental

Research and Public Health, 17(22), 8500.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

486