SPECIAL ISSUE For IJEIT ON ENGINEERING AND INFORMATION TECHNOLOGY. , VOL.12 ,NO.

1, December 2024 41

Design and Implementation of a Mobile E-Commerce

Platform Based-on Machine Learning

Alturki, B. Adabashi, A.
Faculty of Engineering, Sabratha University Faculty of Engineering, Sabratha University
[email protected] [email protected]
https://orcid.org/0000-0002-8339-3836

Abstract— In this paper, the “STORY”, which is an e-

commerce mobile platform, is developed based-on machine Information and communications technology has the
learning techniques in order to consolidate selling methods ability to play an important role in developing and
into a unified, efficient, and personalized environment. Also, improving services for distributing services and solutions
to elevate user experiences, and cultivate a thriving e- in the business sectors Electronic commerce, commonly
commerce environment in Libya. STORY is designed to known as e-commerce, constitutes the exchange of
serve the needs of both customers and merchants, aiming to products or services facilitated through computer
bridge the gap between consumer demands and efficient networks, prominently the Internet. This transformative
business interactions. This platform strives to provide mode of trade has experienced an exponential surge in
customers with a wide range of services and functions that
match their preferences, while empowering merchants with
popularity, aligning with the swift and expansive growth
a user-friendly mobile interface for effective business of the Internet's reach and capabilities. The ascent of e-
management. Among its notable features, STORY boasts an commerce, including both Business-to-Consumer (B2C)
image-based search function, which simplifies the process of and Business-to-Business (B2B) interactions, has
searching for desired or similar products. Additionally, the remarkably surged to the forefront of contemporary
platform integrates an intelligent recommendation system business strategies [2]. This pivotal shift is entrenched in
capable of tracking and analyzing customer behaviors the increasing demand for online services, compounded
within the platform. This system collects pertinent data to further by the indispensable role of technology in catering
offer personalized product recommendations and tailored to the needs of modern consumers. The integration of the
offers, elevating the shopping experience. The system
underwent an evaluation via a questionnaire, which had
internet into the fabric of daily life is nothing short of
overwhelmingly positive results. User feedback and remarkable, according to Statista [3], in January 2024,
satisfaction regarding the system's functionalities and there were 5.35 billion (66.2 percent of the global
performance were notably favorable. population), internet users worldwide.

Index Terms— e-commerce, mobile application, machine The e-commerce domain serves as the nexus where
learning techniques, recommendation system, image-based innovation, convenience, and consumer preferences
search. converge. In the B2C sphere, direct sales from businesses
to end consumers redefine retail paradigms, while B2B
I. INTRODUCTION transactions underpin inter-company commerce.
Irrespective of the domain, the overarching objective

I
nformation and communications technology has
evolved over time, which has the ability to play an
important role in developing and improving services for
remains consistent, facilitating seamless access to goods
and services across diverse digital platforms,
transcending geographical constraints and temporal
distributing services and solutions in the business sectors limitations.
simplifying daily tasks [1]. Nowadays, finding everything
one needs remotely has become effortlessly achievable Moreover, with the widespread use of smartphones,
with a few clicks, provided there's a reliable internet mobile e-commerce has become an integral part of our
connection. As in all other spheres of life, the commercial daily lives. An e-commerce mobile application is a digital
field has also evolved in tandem with technological platform specifically designed and developed for
advancement. With electronic commerce, people may handheld devices that facilitates online buying and selling
now purchase and obtain the things they need without of goods or services.
ever leaving their houses. Of the various technological advancements shaping
‫ـــــــــــــــــــ ـــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــ‬ this landscape, Machine Learning (ML) stands at the
Received 15 Apr , 2024; revised 12 May, 2024; accepted 15 Mar 2024.
forefront, exerting a profound influence on the dynamics
Available online 08 Aug, 2024.
of e-commerce. The application of machine learning

www.ijeit.misuratau.edu.ly ISSN 2410-4256 Paper ID: IT021

 42 Alturki, B. and Adabashi, A./ Design and Implementation of a Mobile E-Commerce Platform Based-on Machine Learning

technologies has redefined the consumer experience. II. LITERATURE REVIEW

Through advanced systems and recommendation engines,
machine learning not only facilitates seamless A. Mobile E-commerce Applications
transactions but also enhances immersive engagements, Electronic commerce or e-commerce is the term used
catering to the diverse needs of an increasingly digital to describe any economic activity, such as selling, buying
consumer base. or exchanging products and services over the Internet [4].
Within the context of Libya, the landscape of Accordingly, mobile e-commerce or m-commerce can be
electronic commerce predominantly relies on social defined as an extension of e-commerce, as the two terms
media platforms. Merchants utilize these platforms to are similar to each other. They both share basic business
showcase their products, initiating customer orders principles. More specifically, m-commerce is just one of
through phone calls or messaging channels. This method the many subsets of e-commerce in which all the
necessitates the collection of customers' personal transactions are connected via handheld devices such as
information each time they make an order for subsequent mobile phones, and their interaction occurs in a wireless
communication regarding delivery arrangements. mode.
Additionally, a secondary approach involves the Recent developments in digital payment infrastructure,
development of institution-specific applications to exhibit coupled with the advent of smartphones and the Internet,
their product catalog. have spurred the mobile commerce revolution around the
In the Libyan market, a limited number of e-commerce world. Mobile e-commerce applications allow online
mobile applications are present. Registering for these sellers to offer personalized shopping experiences
platforms entails a sequence of procedures and through easy-to-use features. The benefits of personalized
contractual arrangements between the application's offers, and faster shopping experiences have led online
operating company and sellers interested in joining the shoppers to prefer mobile applications over the other
platform. This limited accessibility has resulted in online shopping ways [5].
ineffective interaction between the system and customers. B. Review on Existing E-commerce Applications
Consequently, customers lack personalized experiences
as their specific requirements and interests remain The shopping landscape has seen a major shift due to
unspecified. Most applications employ a simplistic the surge of online stores. This review delves into some
keyword-based search method, limiting product online shopping mobile apps in Libya that have been
discovery solely to product or brand names. developed during the last years, exploring the dynamic
world of e-commerce applications, seeking to unravel
The prevailing e-commerce landscape in Libya their intricacies and user experiences. The review
encounters multifaceted challenges. The reliance on highlights four established e-commerce applications that
social media as the primary transaction medium lacks have gained recognition in the Libyan market.
standardization, resulting in disparate processes and
impeding the establishment of a unified and streamlined • Jibli-Mall
electronic marketplace. This fragmentation not only Jibli Mall is the pioneer in Libya, introducing
hampers user experiences but also restricts scalability and comprehensive Quick Commerce Super App that has
accessibility. As a result, there's an urgent need for an transformed the way people shop. Jibli-Mall offers a
integrated and standardized e-commerce platform tailored convenient and seamless solution for purchasing essential
to Libya's market. products whenever they are needed. The app is accessible
Therefore, this study is interested in designing and on both Android and iOS platforms and guarantees
implementing a mobile e-commerce platform called prompt and dependable delivery to all areas of Tripoli,
STORY, which aimed to consolidate various selling Libya. This aligns with the description provided on their
methods into a unified, efficient, and personalized website (www.jibli.ly).
experience. STORY seeks to overcome the limitations of • Spiza
the current landscape by facilitating enhanced
interactions, personalized experiences, and advanced The Spiza application (www.spiza.ly) was established
search features. By centralizing disparate selling practices in 2018 and has been downloaded by 50,000 users,
into one platform, STORY endeavors to streamline according to the company's statistics. This application
processes, provide customers with a wide array of serves as an intermediary between sellers and customers
services and functionalities that match their preferences, who are registered in the program. Presently, Spiza
elevate user experiences, and cultivate a thriving mobile operates exclusively within the city of Benghazi. To
e-commerce environment in Libya. include a business activity within the Spiza program,
individuals or businesses need to submit a request
Among its standout features, STORY boasts an image- through the company's website. The request undergoes a
based search function, simplifying the quest for desired or review process, which includes communication with the
similar products. Additionally, the platform integrates an seller and the preparation of a contract to activate the
intelligent recommendation system capable of tracking business activity. To make a purchase through this
and analyzing customer behaviors both within the program, users must first log in. Registration necessitates
platform. Finally, tested the e-commerce system, and providing a phone number and a delivery address. Once
customer satisfaction is high. logged in, users can place orders, track order status, and
access order details through the ordering interface.

www.ijeit.misuratau.edu.ly ISSN 2410-4256 Paper ID: IT021

SPECIAL ISSUE For IJEIT ON ENGINEERING AND INFORMATION TECHNOLOGY. , VOL.12 ,NO. 1, December 2024 43
• E-zad aims to transcend conventional text-based searches,
enabling users to explore products effortlessly through
This company offers sales and delivery services
visual input. Secondly, the study of recommendation
through its application, which hosts multiple stores.
engines seeks to personalize user experiences by
Additionally, the company has established agreements
suggesting products aligned with individual preferences
with numerous local suppliers and wholesalers. To
and browsing history. Lastly, the utilization of a
become a registered store within the application,
background remover for images aims to elevate product
individuals or businesses need to engage in a personal
visuals, enhancing their appeal and presentation.
contract with the company and create a commercial
agreement, enabling them to showcase their products Search by image, commonly known as reverse image
within the program. Customers can make purchases from search, is an innovative and powerful technology that
the program by choosing from the variety of stores it enables users to uncover information, find related images,
offers, with the assurance of a swift delivery service and access content by using an image as the query instead
(www.elzad.ly). of traditional text-based input. In conventional text-based
search engines, users input keywords to retrieve relevant
• Click-shop
results. However, with image search, users can upload or
Click Shop is a prominent company specializing in provide an image, and the search engine responds by
connecting customers with various restaurants and stores presenting results that encompass visually similar or
through the convenience of the internet. It offers a user- related images, websites, products, and a diverse array of
friendly and efficient way for customers to place orders. content. This transformative approach has redefined how
Whether you're using a personal computer, mobile phone, users navigate and interact with digital information.
or tablet, you can easily make your selection through
Content-based image retrieval (CBIR) is a technique
Click Shop. The process is straightforward: once a
used for retrieve similar images from a database. The
customer submits an order, Click Shop's program swiftly
most challenging aspect of CBIR is to bridge the gap
relays the order to the respective restaurant or store. The
between the low-level feature layout and high-level
company takes the responsibility of ensuring the order
semantic concepts [6]. Convolutional Neural Network
reaches its destination on time. For merchants interested
(CNN) is a machine learning method which has been
in joining the app, registration is possible by submitting
successfully applied to the field of image classification.
an application through the company's website dedicated
The structure roughly mimics the nature of the
to the app (www.clickshoop.net).
mammalian visual cortex and neural networks in the
The following comparison table presents an analysis of brain. It is inspired by the human visual system because
four applications in the Commerce and delivery service of its ability to recognize and localize objects within
industry. This comparative evaluation aims to provide a cluttered scenes [7]. The evolution of image retrieval
comprehensive overview of these applications' key methodologies has seen remarkable strides, notably
features. By examining the strengths and weaknesses of propelled by groundbreaking contributions in the field.
these services. Babenko and Lempitsky [8] introduced a groundbreaking
TABLE 1 REVIEW ON EXISTING E-COMMERCE APPLICATIONS
approach in the domain of image retrieval, unveiling a
novel descriptor named SPoC (sum-pooled convolutional
features), heralding a remarkable leap in retrieval
Recommendation

Chat Features
Personalized

accuracy compared to preceding deep feature-based

Applications

Search Type
Suggestions

Platform
System

descriptors. It distinctly emphasizes the necessity to

reevaluate established computer vision techniques when
applied to deep convolutional features, elucidating the
unique characteristics inherent in these descriptors
Jibli- Android, compared to conventional ones like SIFT (Scale-Invariant
No No Keyword-based No
Mall iOS Feature Transform).
Android,
Spiza No No Keyword-based No
iOS Recommendation engine is a type of data filtering
Keyword-based Android, tool that uses machine learning algorithms to provide
E-zad Yes No No
And QR code iOS personalized recommendations to users based on their
Click- Android,
shop
Yes No Keyword-based
iOS
No preferences and past behaviors. In realm of e-commerce,
recommendation systems are used to support consumer
C. Machine Learning in a Mobile E-commerce choice in the e-commerce applications. Such systems are
In the constantly evolving realm of e-commerce, developed to predict a set of items that may be interesting
mobile devices have emerged as a primary platform for to a specific customer in the absence of information about
online shopping. This shift towards mobile has created the choice of specified product by this user. In making
opportunities for businesses to leverage machine learning recommendations data on the choice of other users, as
techniques in improving the shopping experience, well as on the characteristics of selected items are used
tailoring interactions, and boosting sales. As a subset of [9]. In the dynamic and rapidly evolving landscape of e-
artificial intelligence, ML empowers computers to learn commerce, characterized by an abundance of choices that
from data, rendering it a potent instrument for optimizing confront online shoppers, recommendation systems
strategies in mobile e-commerce. assume the role of indispensable digital shopping
companions. These systems constitute the very essence of
This research paper delves into three ML applications. personalized shopping experiences. They play a pivotal
Firstly, the investigation into image search functionalities role in assisting users as they navigate the expansive

www.ijeit.misuratau.edu.ly ISSN 2410-4256 Paper ID: IT021

 44 Alturki, B. and Adabashi, A./ Design and Implementation of a Mobile E-Commerce Platform Based-on Machine Learning

online marketplace, enabling them to make well-informed Inspired by these applications, this paper presents the
choices with confidence. design and implementation of a mobile e-commerce
system using machine learning techniques.
Over the years, several different approaches for
generating recommendations have been proposed. Bansal
et al. [10] applied GRU-based recurrent neural networks III. METHODOLOGY
to transform item text into latent features. This approach The Software Development Life Cycle known as the
aimed to enhance collaborative filtering performance, Iterative Waterfall Model was used for developing this
particularly for addressing cold start challenges. They mobile e-commerce application. This approach allows for
tested the model on a citation recommendation system periodic reviews and adjustments in each phase of
using real-world datasets from CiteULike, including both development, offering greater flexibility and adaptability
dense and sparse versions. Their model was compared to changing requirements and stakeholder feedback while
with a modified version of collaborative topic modeling. still maintaining the structured, sequential nature of the
Across both datasets, their proposed model demonstrated Waterfall Model. The typical phases in this model include
statistically significant performance improvements. requirements, design, implementation and testing. The
Zheng et al. [11] presented DeepCoNN (Deep details of each phase are given in the following
Cooperative Neural Networks) as a novel approach to subsections.
recommender systems. DeepCoNN applies deep learning A. Requirements
to jointly capture user and item characteristics using
reviews. It incorporates two parallel neural networks— In the pivotal phase of requirements gathering, a
one for learning user preferences and another for item comprehensive exploration of the research objectives and
properties from reviews. These networks share a top stakeholder requirements is systematically conducted.
layer, allowing interaction between learned factors for This research used two methods to collect requirements:
users and items, resembling factorization machines. competitive analysis and market research.

Liang and Baldwin [12] introduced a novel Gathering requirements through competitive analysis is
recommender system approach called the Probabilistic a valuable method for comprehending market trends,
Rating Auto-Encoder. This system leverages customer expectations, and industry best practices.
unsupervised feature learning and generates user profiles Through the examination of competitors and their
from user-item rating data to enhance collaborative offerings, one can derive invaluable insights to inform the
filtering methods. By adding deep learning to development of their own products or services. The
conventional approaches like matrix factorization, advantages of competitive analysis include gaining a
particularly with the yelp.com dataset in the e-commerce deeper understanding of the market landscape, learning
domain, they observed statistically significant about customer preferences and expectations.
improvements in rating prediction. Market research encompasses the collection, analysis,
V. Paranjape et al. [13] deployed a machine learning and interpretation of data related to the market and
approach for item recommendation in e-commerce potential customer base. Through meticulous
making use of collaborative based model. This research examination, it enables organizations to acquire crucial
follows the model based approach using the concept of insights that inform strategic decisions. The advantages
singular value decomposition (SVD) in order to get the of market research include preventing the wastage of
ratings of unrated items and minimize the sparsity resources on ineffective strategies, understanding
problem. customer preferences and needs, and proactively
addressing potential risks and challenges.
Image background remover, in the realm of image
processing, a machine learning-based background The most important aspect of obtaining a
remover stands as an advanced technique utilizing deep comprehensive picture of the interactions between users
learning algorithms, particularly convolutional neural and the platform is to perform a system requirements
networks. Its primary function is to automatically identify analysis. The paper employs the use of use case diagrams
and isolate foreground objects from their backgrounds to model user interactions with system components. This
within digital images. The core principle revolves around serves the purpose of identifying and elucidating user
the algorithm's ability to recognize patterns, textures, and requirements. The main functions of the system, as
contours, enabling it to accurately delineate the depicted in Figure 1, represents the core operations and
foreground objects and remove the surrounding activities that the system is designed to carry out. These
background. functions hold critical importance, as they are essential in
guaranteeing the system's effective realization of its
intended goals and in meeting the needs of its users.

www.ijeit.misuratau.edu.ly ISSN 2410-4256 Paper ID: IT021

SPECIAL ISSUE For IJEIT ON ENGINEERING AND INFORMATION TECHNOLOGY. , VOL.12 ,NO. 1, December 2024 45
Categories, product classification and associated stores;
Carts for temporary storage of items specified by
customers; Wish lists that allow users to save desired
products; orders, and record purchase details.
C. Implementation
The implementation phase marks the transformative
transition from abstract concepts and meticulous planning
to the tangible realization of the project. It delineates the
critical phases of implementation, pivotal in translating
design into a functional software solution while ensuring
its reliability and compliance with specified requirements.
It involves writing the actual code, designing the user
interface, structuring the database, and interconnecting all
the components.
Implementing the designed system requires
consideration of two technical parts, namely
programming languages and database.
Programming Languages: the primary programming
language employed in this research was Dart,
FIGURE 1. MAIN FUNCTIONS OF THE SYSTEM empowering the construction of not only the backend
logic but also the mobile frontend component. Its
B. System Design versatility allowed for the seamless integration of the core
This paper aims to develop and design a mobile e- backend functionalities with the mobile user interface.
commerce platform application named STORY. Figure 2 PHP which stands for Hypertext Preprocessor, is a
describes the system. versatile and widely adopted scripting language primarily
The mobile application for e-commerce is comprised used for web development. It plays a pivotal role in
of essential sections that cater to different functions. To enabling dynamic and interactive web applications. PHP
ensure the safety of users, User Authentication module was leveraged as the intermediary layer to establish a
authenticates their login and registration details. The robust connection with the database, enabling efficient
Product Listing feature organizes all available products, data retrieval, manipulation, and storage operations.
while the Image Processing module optimizes visuals for Python is a versatile and widely acclaimed
an appealing display. Easy item discovery is facilitated by programming language known for its simplicity,
the Product Search module. Customer Messaging module readability, and robust capabilities. It has gained immense
caters to user-support interactions. popularity across various domains, from web
development and data science to artificial intelligence and
automation. Python was employed for the implementation
of ML algorithms as a crucial component of this work.
Python's extensive libraries and frameworks dedicated to
ML facilitated the development of intelligent and data-
driven functionalities, enhancing the overall system's
capabilities.
Node.js is a powerful, open-source JavaScript runtime
built on Chrome's V8 JavaScript engine. It allows
developers to execute JavaScript code outside of a web
browser, making it an exceptional choice for building
server-side applications and networked services. Node.js
was utilized in the development of a real-time messaging
FIGURE 2. SYSTEM MODULES server.
Furthermore, seamless purchase processes are Database: two type of database were used in this
guaranteed with Shopping Cart and Checkout module work, MySql and MongoDB. MySQL is a widely used
complemented by Recommendation Engine module open-source relational database management system that
providing personalized recommendations. With Order plays a fundamental role in managing and storing
Management handling processing activities as well as structured data. MySQL was used for managing the
tracking orders, Analytics and Reporting module offers relationships of basic data entities in the system including
valuable insights useful in decision-making processes. customers, products, categories, orders and wish lists.
Also this paper makes use of a NoSQL database,
The main entities of the system are: Customers, which
specifically MongoDB, which is a popular and versatile
contain the personal details of registered users; Stores,
database designed to efficiently handle unstructured or
which represents seller information and store-specific
semi-structured data. MongoDB complements the
attributes; Products, and item descriptions available;
system's relational database by specifically providing

www.ijeit.misuratau.edu.ly ISSN 2410-4256 Paper ID: IT021

 46 Alturki, B. and Adabashi, A./ Design and Implementation of a Mobile E-Commerce Platform Based-on Machine Learning

real-time chat services and capturing user system of customers in the realm of buying and selling.
interactions. In MongoDB, data is organized into Individuals have the option to register on this platform,
collections, which are similar to tables in relational either as merchants to showcase their products or as
databases. Each collection contains a set of documents, customers to peruse and make purchases.
where each document represents an individual item or
The application is accompanied by a dedicated
entity.
website shown in Figure 3 that provides detailed insights
Machine learning techniques: In the realm of into the application's features and usage guidelines. User
machine learning implementation, a variety of techniques can easily access this website to download the
were used. Singular Value Decomposition (SVD) was application.
utilized for collaborative recommendations, while TF- FIGURE 3. STORY WEBSITE
IDF and cosine similarity integration for content-based
recommendations. For image feature extraction, The major objectives of developing STORY
ResNet80 Convolutional Neural Network (CNN) was
utilized for both image search and content
recommendation, and incorporated ANNOY as the search
algorithm. Finally, the extraction of the image's
background was successfully accomplished through the
utilization of Rembg algorithm. The Rembg algorithm is
an advanced tool based on deep learning, specifically
utilizing the U-2-Net-L neural network architecture to
precisely remove backgrounds from images. This
algorithm operates on the principle of semantic
segmentation, where it identifies and understands the
semantic meaning of different parts of an image.
Furthermore, in this study, Socket.IO a JavaScript
library, was used to enable real-time, bidirectional
communication between clients (users and devices) and
the system servers. It utilizes WebSocket as the
underlying protocol to provide low-latency, full-duplex
communication channels. To utilize Socket.IO, the initial
step involves incorporating the Socket.IO library on both
the client and server facets of the application. This
ensures seamless communication between the client and
server, facilitated by Socket.IO's protocol.
D. Testing application are to:
 Provide customers with product recommendations
The testing phase ensures the integrity and quality of based on their behavior and individual preferences to
the system by thoroughly examine the system to uncover enhances their shopping experience.
and resolve any issues or irregularities. This phase is like
a final check to make sure the system functions smoothly  Provide customers multiple search options, including
and meets the intended requirements. Various types of image-based search, and text search to diversifies
tests were conducted to address every aspect of the their browsing experience.
mobile e-commerce application.  Provide tools and interfaces for vendors to manage
Unit testing involves testing each function in the their product listings, inventory, pricing, and order
system individually, isolating it from its dependencies to fulfillment efficiently.
ensure the excellence and correctness of each function.  Streamline the order processing and fulfillment
Integration testing is a multifaceted process that extends process, including order tracking and communication
to data integrity. This phase is crucial in ensuring that between vendors and customers.
data is seamlessly shared and processed between various
 Offers sellers the opportunity to deal directly with
components, guaranteeing the integrity and reliability of
customers via an internal messaging system
the application.
integrated into the platform, which simplifies the
Furthermore, system restrictions and permissions process of inquiries and negotiations.
testing was performed, which is an essential step to  Enables sellers to effectively remove the background
ensure that the system accurately enforces access from product images, thereby enhancing their visual
restrictions and permissions assigned to different user appeal and overall attractiveness.
roles.
 Provide vendors with analytics tools to track sales,
customer behavior, and other key metrics, enabling
IV. THE PLATEFORM
data-driven decisions.
The STORY platform developed in this research
represents a system designed to cater to the requirements

www.ijeit.misuratau.edu.ly ISSN 2410-4256 Paper ID: IT021

SPECIAL ISSUE For IJEIT ON ENGINEERING AND INFORMATION TECHNOLOGY. , VOL.12 ,NO. 1, December 2024 47

 Ensure a user-friendly, seamless shopping experience

for customers.
As mentioned earlier, the user can be a customer or a
seller, so the application offers two user interfaces, the
customer interface and the seller interface.

B. Seller Interface
The Seller homepage consists of the statistics, orders,
financial analytics and customers section as shown in
Figure 7. The statistics section (‫ )االحصبئيبت‬provides
A. Customer Interface sellers with crucial insights and metrics related to their
performance, order volumes, customer acquisition, and
FIGURE 4. GENERAL CATEGORIES PAGE SCREENSHOT daily sales. The Order Analytics section (‫ )الطلببت‬provides
sellers with detailed insights into specific order statuses,
The customer interfaces serve as the primary analytical drawing displaying the percentage of orders
interaction point for users browsing and purchasing attributed to each category, individual product sales, and
products. The customer homepage consists of the sales frequency over different time frames.
recommendation product tab, followed store, search, and
the general categories page. The general categories page The Financial Analytics section (‫)البيبنبت المبلية‬
(‫ )الفئبت‬shown in Figure 4, showcases a structured layout visualizes the information about total sales, profits, and
presenting various product categories available within the profit percentages through bar charts that depict sales and
application. It enables users to browse through diverse profits over different time periods. In addition, it provides
categories, offering a clear and organized navigation path. graphs to compare weekly, monthly and yearly sales
against previous periods. Furthermore, it provides
insights into average order value and breaks down sales
figures by product category.
The customer section (‫ )العمالء‬offers sellers insights
into customer acquisition trends, order frequency, and a
breakdown of customers based on their order count
within various time frames.

FIGURE 7. SELLER HOMEPAGE SCREENSHOT

V. SYSTEM EVALUATION
FIGURE 5. RECOMMENDATION PAGE SCREENSHOT
In the process of evaluating the system, the
The recommendation product page (‫ )لك‬described in questionnaire consisting of 12 questions was used. It aims
Figure 5, offers personalized recommendations tailored to to gather insights into the user experience, identify areas
individual user preferences. The followed store tab of improvement, and assess overall satisfaction. Each
(‫ )المتببعة‬exclusively showcases products from stores question within the questionnaire has been carefully
followed by the user. Figure 6 shows the search tab selected and worded with meticulous precision to
(‫ )البحث‬offers versatile search capabilities, empowering effectively measure usability. Questions related to
users to explore products through both text-based queries usability cover various aspects, including ease of
and image inputs. navigation, task efficiency, and system responsiveness.
FIGURE 6. SEARCH PAGE SCREENSHOT

www.ijeit.misuratau.edu.ly ISSN 2410-4256 Paper ID: IT021

 48 Alturki, B. and Adabashi, A./ Design and Implementation of a Mobile E-Commerce Platform Based-on Machine Learning

FIGURE 8. SURVEY RESPONSES

VI. CONCLUSIONS
This study applied ML techniques to develop a mobile
e-commerce platform called STORY, in order to support
the advancement of the Libya's online retail market. This
platform seamlessly integrates numerous features,
including multi-vendor functionalities and cutting-edge
AI elements such as image search, background removal,
recommendation systems, real-time chat, and
comprehensive vendor analysis tools. Furthermore, the
utilize of open-source technologies such as PHP,
MySQL, Dart, Node.js, and Python not only underscores
the application's robustness, but also emphasizes its
commitment to simplicity, performance, and security.
Moreover, the evaluation using questionnaires
highlighted its efficacy in enhancing user experiences and
empowering vendors with data-driven insights.
For future research, this study suggests to add some
The process begins with the collecting basic features to increase the application performance and
demographic data from the participants. Followed by users’ satisfaction such as:
questions evaluating ease of use. These questions are  Implement a range of electronic payment options,
specifically designed to assess the ease of use, including digital wallets, mobile payment systems,
satisfaction, and system efficiency. They are tailored to and additional emerging technologies, to offer users
provide insights into the user's perception of how user- with diverse and secure methods to conduct in-app
friendly and convenient the software is in practical use. transactions.
The questionnaire concludes with an open-ended  Develop an integrated delivery service program
question, inviting participants to provide qualitative within the application framework and orders tracking
comments and express their thoughts using their own to ensure efficient and reliable delivery experiences
words. This qualitative data collection method gathers for purchased items.
deeper insights into participants' experiences with the
 Develop a feature allowing users to seamlessly
system.
convert product images into interactive 3D models
The distribution strategy for collecting user feedback within the application.
involved utilizing the Facebook platform within a
dedicated group over a period of few days. REFERENCES
The questionnaire survey, drawing 20 responses, [1] Asif Rahman, Md., M. A., and Khan, Rabiul Islam Md., 2022.
reflected a gender distribution of 33% male and 66% Design and Implementation of a Web-Based Electronic-
female. The user satisfaction resonated positively, with Commerce System. European Journal of Engineering and
Technology Research, Vol 7.
38% expressing being "very satisfied" and 53% reporting [2] DOI: http://dx.doi.org/10.24018/ejeng.2022.7.2.2789
as "satisfied" while a remaining 9% remained "neutral" [3] Meesam, M. 2021. "Business Organization and e-Commerce".
Regarding usability, a substantial 70% rated it as "very Journal of Advance Research in Science and Social Science
good" with an additional 25% considering it "good". (JARSSC) 4(1),pp. 181-191.
[4] Statista:https://www.statista.com/statistics/617136/digital-
These figures indicate that a commendable user population-worldwide/, accessed on Fepruary 11, 2024.
experience and high usability. [5] Niranjanamurthy, M., Kavyashree, N., Jagannath , S., and Chahar,
D. 2013. ―Analysis of ecommerce and m-commerce: advantages,
limitations and security issues,‖ International Journal of Advanced
Research in Computer and Communication Engineering, 2(6),
2360- 2370.
[6] Chopdar, P.K., Balakrishnan, J. 2020. ―Consumers response
towards mobile commerce applications: SOR approach,‖
International Journal of Information Management, 53, 102106.
[7] Alzu’bi, A., Amira, Ramzan, A. N. 2015. ―Semantic Content-
based Image Retrieval: A Comprehensive Study,‖ J. Vis.
Commun. Image R.
System Usability
[8] Krizhevsky, A., Sutskever, I., and Hinton, G. 2017. ―Imagenet
classification with deep convolutional neural networks,‖ Commun.
ACM, vol.60, pp. 84–90.
[9] Babenko, A., and Lempitsky, V. 2015. ―Aggregating local deep
features for image retrieval. In International Conference on
Computer Vision (ICCV)‖, Proceedings of the IEEE International
Conference on Computer Vision (ICCV), pp. 1269-1277.
[10] R. Burke, ―Hybrid web recommender systems. In: The adaptive
web‖, pp. 377–408, 2007.

User satisfaction

www.ijeit.misuratau.edu.ly ISSN 2410-4256 Paper ID: IT021

SPECIAL ISSUE For IJEIT ON ENGINEERING AND INFORMATION TECHNOLOGY. , VOL.12 ,NO. 1, December 2024 49
[11] Bansal, T., Belanger, D., and McCallum, A. 2016. ―Ask the gru:
Multi-task learning for deep text recommendations,‖ In
Proceedings of the 10th ACM Conference on Recommender
Systems.
[12] Zheng, L., Noroozi, V., Yu, P. S.. 2017. ―Joint deep modeling of
users and items using reviews for recommendation,‖ Proceedings
of the tenth ACM international conference on web search and data
mining.
[13] Liang, H., and Baldwin, T. 2015. ―A probabilistic rating auto-
encoder for personalized recommender systems,‖ Proceedings of
the 24th ACM International on conference on information and
knowledge management.
[14] Paranjape,V., Nihalani, N., and Mishra, N.2021. "A Machine
Learning Approach for Item Recommendation Using SVD
Technique," IEEE 4th International Conference on Computing,
Power and Communication Technologies (GUCON), Kuala
Lumpur, Malaysia, pp. 1-7.

www.ijeit.misuratau.edu.ly ISSN 2410-4256 Paper ID: IT021