JOURNAL INFORMATIC.
EDUCATION AND MANAGEMENT (JIEM)
Vol 8 No 1 .
: September 2025 - February 2026, pp.
ISSN: 2716-0696.
DOI: 10.
61992/jiem.
A 636
Analysis of E-Commerce and Fintech Trends in the Digital Economy Ecosystem Redha Bayu Anggara 1*.
Asyrof Fitrah 1.
Ali Ibrahim 1.
Mira Afrina 1 Universitas Sriwijaya Article Info Article history:
Received 31 December 2025 Revised 1 January 2025 Accepted 2 January 2026 Keywords:
Ecommerce.
Fintech
ABSTRACT
The rapid expansion of e-commerce and fintech has significantly shaped the digital economy ecosystem in Indonesia.
This study analyzes key trends, behavioral patterns, and the evolving dynamics within these sectors as digital adoption continues to accelerate.
The increasing volume and complexity of digital transactions demand advanced analytical approaches capable of identifying hidden patterns and potential To address this need, the study employs a Convolutional Neural Network (CNN) model to extract deep feature representations from transaction data and classify emerging behavioral trends.
The proposed method demonstrates strong accuracy, achieving 91.
25%, indicating its ability to capture non-linear relationships that traditional methods often The findings highlight several major trends, including shifting consumer behavior, increasing transaction frequency, and the growing prominence of digital financial services.
Practically, this research provides valuable insights for enhancing risk mitigation, fraud detection, and real-time monitoring in digital platforms.
Academically, it contributes to the understanding of deep learning applications in digital economic analysis and opens avenues for further research on hybrid models and multi-source data integration within the digital economy ecosystem.
This is an open access article under the CC BY-SA license.
Corresponding Author:
Redha Bayu Anggara | Universitas Sriwijaya Email: redhabayuanggara@gmail.
Introduction The rapid expansion of digital technologies has reshaped global economic activities, with ecommerce and financial technology .
emerging as key drivers of digital These sectors have enabled seamless online transactions, enhanced business efficiency, and broadened access to financial services.
In developing countries such as Indonesia, the adoption of digital platforms has accelerated rapidly, influencing consumer behavior, business models, and overall market dynamics within the digital economy.
Journal homepage: http://w.
id/index.
php/jiem JOURNAL INFORMATIC.
EDUCATION AND MANAGEMENT (JIEM) Vol 8 No 1 .
: September 2025 - February 2026, pp.
ISSN: 2716-0696.
DOI: 10.
61992/jiem.
A 637
As digital interactions grow, the volume and complexity of transaction data continue to rise.
This growth presents both opportunities and challenges for policymakers, businesses, and On one hand, rich datasets provide valuable insights into market trends, consumer preferences, and economic performance.
On the other hand, the surge in data complexity brings increased risks, including transaction anomalies, cybersecurity threats, and online fraud.
Understanding these patterns requires systematic analytical approaches capable of capturing evolving behavioral and market trends.
Research on the digital economy has highlighted the importance of analyzing user activity, transaction flows, and marketplace dynamics to support decision-making and strengthen digital ecosystem resilience.
Identifying emerging trends in e-commerce and fintech not only helps organizations improve their strategic planning but also supports governments in shaping policies for sustainable digital economic growth.
This study aims to examine the key trends influencing the e-commerce and fintech landscape and to provide a comprehensive overview of how digital behaviors evolve within the broader digital economy ecosystem.
By leveraging publicly available datasets and structured analytical methods, this research offers insights that can support the development of more secure, adaptive, and innovation-driven digital economic environments.
Research Methodology Data Source The research data was obtained from a public Kaggle dataset containing more than 1 million e-commerce and fintech transaction entries.
The dataset includes variables such as Transaction Amount.
Payment Method.
Device Type.
Timestamp.
User Age, and Is Fraudulent.
Analysis Stages Data Preprocessing Oe Removing duplicate records and missing entries.
Oe Applying normalization using RobustScaler to reduce the influence of extreme Feature Engineering Oe Transforming timestamp variables into daily and weekly behavioral cycle patterns.
Oe Adding behavioral attributes such as transaction frequency, average purchase value, and IP distance between users.
Feature Extraction and Classification (CNN) Oe Numerical data are converted into matrix-like representations .
Oe The Convolutional Neural Network (CNN) extracts spatial and temporal features from these matrices.
Oe The final dense layers of CNN perform classification to detect transaction patterns and anomalies.
Journal homepage: http://w.
id/index.
php/jiem JOURNAL INFORMATIC.
EDUCATION AND MANAGEMENT (JIEM) Vol 8 No 1 .
: September 2025 - February 2026, pp.
ISSN: 2716-0696.
DOI: 10.
61992/jiem.
A 638
Model Evaluation Oe Evaluation metrics include Accuracy.
Precision.
Recall.
F1-Score, and ROC-AUC.
Oe Model performance is compared against baseline methods such as Logistic Regression and Random Forest.
Results and Discussion Experiment Setup and Results The dataset was split into 80% training, 10% validation, and 10% testing.
A Convolutional Neural Network (CNN) was employed to extract behavioral patterns from user transactions and to perform classification of normal and anomalous activities.
Model training was carried out using Python (TensorFlo.
on an RTX-series GPU to accelerate convergence.
Table 1.
Performance Comparison of Models Model
Accuracy Precision Recall F1-Score
ROC-AUC
Logistic Regression Random
Forest
CNN
Performance Analysis The CNN model demonstrates a substantial improvement compared to conventional By extracting non-linear and spatial relationships from transactional behavior.
CNN provides a more accurate and stable performance without signs of overfitting.
The deep feature extraction process allows the model to capture subtle behavioral anomalies that traditional algorithms tend to miss.
Trend Analysis The experiment reveals several significant behavioral patterns within the digital economy E-commerce: Transaction volume increases by 24% during weekends, indicating a strong consumer shopping preference at the end of the week.
Fintech: A 32% spike occurs at the beginning of each month, driven by salary cycles, e-wallet top-ups, and micro-loan activities.
Fraud Pattern: Most anomalies occur in cross-regional transactions involving new devices and credit card payments, suggesting duplicated accounts or suspicious identity patterns.
Journal homepage: http://w.
id/index.
php/jiem JOURNAL INFORMATIC.
EDUCATION AND MANAGEMENT (JIEM) Vol 8 No 1 .
: September 2025 - February 2026, pp.
ISSN: 2716-0696.
DOI: 10.
61992/jiem.
A 639
Table 2.
Traffic Transaction Year E-Commerce Value (T IDR) Fintech Users (M) Implications and Discussion The CNN approach proves highly effective not only for anomaly detection but also for uncovering meaningful digital economic insights.
Deep learning provides a robust framework for early warning systems in digital risk monitoring, particularly due to its ability to recognize hidden behavioral structures and complex data interactions.
Model Performance Visualization The following graphs illustrate the difference in model performance during training and Accuracy per Epoch CNN: Accuracy improves steadily from 82% Ie 91% over training epochs, showing stable learning progression without divergence.
ROC-AUC Curve The CNN achieves a ROC-AUC score of 0.
94, indicating strong capability in distinguishing between normal and abnormal .
These visualizations confirm that the model maintains stability during training and performs reliably on unseen data.
Trend Analysis of E-Commerce and Fintech Further exploration of the classified results and graphical trend visualization highlights:
E-commerce: The highest transaction frequencies occur on weekends (SaturdayAe Sunda.
, with a 24% increase compared to weekdays.
Fintech: Early-month activity shows a 32% rise, aligned with salary disbursement, digital wallet top-ups, and loan repayment cycles.
Fraud Pattern: Anomalies frequently appear in cross-regional, card-based transactions initiated from new or unknown devices, indicating potential duplicated accounts or credential misuse.
Journal homepage: http://w.
id/index.
php/jiem JOURNAL INFORMATIC.
EDUCATION AND MANAGEMENT (JIEM) Vol 8 No 1 .
: September 2025 - February 2026, pp.
ISSN: 2716-0696.
DOI: 10.
61992/jiem.
A 640
Figure 1.
Analitis Of Ecomerce and Fintech YoY These findings illustrate how CNN-based analysis can uncover relevant economic insights that support policy-making and market strategies.
Discussion and Interpretation The strong performance of the CNN model demonstrates its ability to capture complex and non-linear behavioral relationships within large-scale digital transaction data.
Traditional models fail to capture these deeper patterns due to their limited feature extraction capability.
From a digital economy perspective, the results reinforce that:
Large-scale transaction analysis benefits significantly from deep learningAebased AI-driven economic analytics can serve as a foundation for digital risk early-warning systems in national financial and e-commerce sectors.
Such methods can be expanded to build dynamic recommendation systems, allowing digital platforms to tailor promotions, pricing, and user segmentation based on real-time behavioral trends.
Research Implications Practical Implications The CNN model can be utilized by:
Oe Fintech and e-commerce providers, for automated anomaly detection, consumer behavior profiling, and customer segmentation.
Oe Financial regulators .
OJK/BI), as a foundation for real-time monitoring of digital transactions.
Oe Digital business operators, to forecast market trends and consumer demand Journal homepage: http://w.
id/index.
php/jiem JOURNAL INFORMATIC.
EDUCATION AND MANAGEMENT (JIEM) Vol 8 No 1 .
: September 2025 - February 2026, pp.
ISSN: 2716-0696.
DOI: 10.
61992/jiem.
A 641
Academic Implications The study contributes to the development of deep learningAebased classification frameworks that can be adapted for various big-data fields.
Conclusions The use of a CNN-based approach significantly improves the accuracy of digital trend analysis achieving 91.
25% accuracy and demonstrates strong capabilities in detecting anomalous transactions with high precision.
This method is scalable and effective for supporting national financial risk prediction systems.
The findings also align with SDG 8 (Decent Work and Economic Growt.
and SDG 9 (Industry.
Innovation, and Infrastructur.
by strengthening sustainable digital transformation.
Acknowledgements The authors would like to express their sincere gratitude to all individuals and institutions that contributed to the completion of this research.
Special appreciation is extended to the academic supervisors and faculty members who provided valuable guidance, constructive feedback, and continuous support throughout the study.
The authors also acknowledge the availability of public datasets and open-source tools that made the experimentation and analysis possible.
Appreciation is given to the computing facilities and resources that enabled efficient model training and evaluation.
Finally, the authors thank their colleagues, peers, and family members for their encouragement, cooperation, and motivation during the research process.
References