Personalized E-commerce based recommendation systems using deep-learning techniques

IAES International Journal of Artificial Intelligence (IJ-AI)
Vol. 13, No. 1, March 2024, pp. 610~618
ISSN: 2252-8938, DOI: 10.11591/ijai.v13.i1.pp610-618  610
Journal homepage: http://ijai.iaescore.com
Personalized E-commerce based recommendation systems using
deep-learning techniques
Shruthi Nagraj, Blessed Prince Palayyan
Department of Computer Science and Engineering, Presidency University, Bangalore, India
Article Info ABSTRACT
Article history:
Received Apr 12, 2022
Revised Apr 26, 2023
Accepted Jul 19, 2023
As technology is surpassing each day, with the variation of personalized drifts
relevant to the explicit behavior of users using the internet. Recommendation
systems use predictive mechanisms like predicting a rating that a customer
could give on a specific item. This establishes a ranked list of items according
to the preferences each user makes concerning exhibiting personalized
recommendations. The existing recommendation techniques are efficient in
systematically creating recommendation techniques. This approach
encounters many challenges such as determining the accuracy, scalability, and
data sparsity. Recently deep learning attains significant research to enhance
the performance to improvise feature specification in learning the efficiency
of retrieving the necessary information as well as a recommendation system
approach. Here, we provide a thorough review of the deep-learning
mechanism focused on the learning-rates-based prediction approach modeled
to articulate the widespread summary for the state-of-art techniques. The
novel techniques ensure the incorporation of innovative perspectives to
pertain to the unique and exciting growth in this field.
Keywords:
Accuracy
Data sparsity
Deep-learning
Recommendation system
Scalability
This is an open access article under the CC BY-SA license.
Corresponding Author:
Shruthi Nagraj
Department of Computer science and Engineering, Presidency University
Bangalore, India
Email: shruthin_12@rediffmail.com
1. INTRODUCTION
Nowadays, the majority of people use social networks and the internet to convey their thoughts,
feelings, and experiences. This typically results in the Internet being used to communicate enormous amounts
of data. Most industrial companies and election campaigns, for instance, rely on understanding the opinions of
people through communication sites and determining whether they are good, negative, or neutral. However,
the majority of this data is beneficial when assessed [1]. The quantity of commodity information is necessary
to considerably improve the commodity information so that users may simply access the product information
they need to avoid information overload. Referral services have become a crucial resource for individualized
information requests in e-commerce as technology develops. There are two phases in the personalized
suggestion service.
Recommender systems are information filtering technologies that are used to address such issues and
give users potentially more relevant and tailored material [2]. Many online systems now include a suggestion
technique, giving internet users more options/services in everyday activities like shopping, listening to music,
and watching movies. The product or service that the system recommends to its consumers is referred to as an
‘item’ in a standard recommender system. While projecting ratings for a single item requires recommender
systems to use the products' descriptions, recommending things to a specific user necessitates recommender
systems that examine the past preferences of users with similar interests. Based on these two categories of

Int J Artif Intell ISSN: 2252-8938 
Personalized E-commerce based recommendation systems using deep-learning techniques (Shruthi Nagraj)
611
methods, recommendation models are segmented into collaborative filtering and content-based filtering.
Hybrid recommendations are an alternative tactic that mixes two or more different kinds of recommendation
algorithms. Recommendation systems [3], [4] achieve commercial success with increasing popularity in a
variety of real-world applications. Such as Walmart, iTunes, and Amazon are online retailers. Adjusting the
suggestions based on a customer's past purchases to include additional goods and services, is widely accepted
for small adjustments to recommendation algorithms to improve the efficiency and financial success of e-
commerce applications. To improve the efficiency of the advice, the e-commerce sector customizes a product's
features for a single consumer. The recommender system should take into account the price as one distinct
property. When a buyer likes a recommended product but decides against it or a rise in the price, personalized
promotion is used to increase the value [5]. A customer may prefer a recommended product but decide against
buying it because of price or other factors. Recommendation system by machine-learning technology. Users
are extremely confused when they get a bulk of information and cannot capture the main points.
The study on recommendation systems is carried out crucially for pricing is highly recognized. To
introduce personalized promotion [6] into recommender systems for e-commerce. Here, the major goal is to
increase the usefulness of product recommendations to personalize the product on a phase-by-phase basis.
Large sets of user data have been generated since the internet's inception. Recommendation systems learn from
the previous history of a customer and recommend products based on their interest [7]. Figure 1 shows the
recommender system for e-commerce. New applications of recommender systems in e-commerce are based on
the examination of interfaces for recommender systems, consumer interfaces, and the technology used to make
suggestions to give inputs they need from the customers.
Figure 1. Recommender system for e-commerce
2. RELATED WORK
With the difficulty associated with the recommendation systems, the most basic issue is to predict
users’ ratings and purchasing views of items based on the prediction. Identify similar users’ interests manually
based on their recommendations of similar users [8]. Manual constructions are largely placed for
recommendation algorithms, like collaborative-filtering mechanisms and content-based filtering mechanisms,
and hybrid algorithms.
Because collaborative filtering algorithms [9] assume that users will have similar preferences for new
items, they will be recommended highly ranked items that are thought to be the most similar to the user's
content. This expands on the items that highlight new products that are comparable to those the user has
previously enjoyed. The fundamental problem is on identifying the user's profile through specific input

 ISSN: 2252-8938
Int J Artif Intell, Vol. 13, No. 1, March 2024: 610-618
612
regarding how much the prior items appeal to the user. The binary classification tasks are solved using a variety
of methods, including support vector machines, K nearest neighbors (K-NN), neural networks, and logistic
regression [10].
A recommendation system for an E-commerce platform utilizing the collaborative-filtering technique
is deployed here for personalized recommendations. A behavior record module is introduced [11], module
analysis is carried out, and a recommendation algorithm module is considered. A personalized recommendation
algorithm is considered for maximum margin factorization used for a personalized recommendation to obtain
results for semi-definite users. A suggested algorithm can deliver users that are more suitable with greater
accuracy than other methods already in use [12].
An efficient technique to enhance the effectiveness of the recommendation by adding the social
network information [13] within the collaborative filtering is carried out by collecting all the data about the
user’s preference ratings along with their social media relationship by extracting data from a social media
networking website. Then to evaluate the performance for collaborative filtering along with all the diverse
neighbor groups comprises of combining groups of friends as well as nearest neighbors.
A novel technique is established for improvising the performance regarding collaborative filtering
recommendations. This is done by collaborating the sparse rating data [14] that is gathered from users along
with the sparse social trust network within the same user base. Stereotyping is one of the first user modeling
and recommendation classes. Rich utilized it for the first time in the Grundy recommender system, which
suggested books. Rich's users were subject to psychological preconceptions that facilitated quick assessments
of persons by researchers [15]. Rich's descriptions of stereotypes, which he referred to as “facets,” are
assemblages of qualities based on these extremely distinctive characteristics. Grundy made the presumption
that male users in this case have “a very strong knowledge of the English language [16], [17],” a high tolerance
for suffering and brutality, a taste for suspense and thrills, and a dislike for slow-moving fiction. Grundy
consequently recommended books. These were thoughtfully arranged according to the facets.
One of the most often used and extensively studied recommendation classes is content-based filtering
(CBF) [12]. The user modeling process, which derives user interests from the objects users interact with, is a
key element of CBF. Typically, “items” are text-based, like emails [18] or web pages [19]. Usually,
“interaction” is established through acts like downloading, purchasing, creating, or labeling anything. A content
model that contains the features of the object is used to represent it. Word-based features, such as single words,
sentences, or n-grams, are the norm. Other non-textual criteria used by certain recommender systems include
writing style, layout details, and extensible markup language (XML) tags [20]. The users and these features
are frequently weighted, and only the most descriptive features are typically employed to model an object.
Combining collaborative filtering and content-based filtering [21], hybrid recommendation
algorithms [10] outperform either filtering process by itself. One major exception found here is transferrable
to other categories to aid prediction in the recommendation system. Customers can express preferences for
introducing new customer items when there are new products available, enhancing turnover. Information
retrieval using this technique is being studied in a variety of industries, including distance learning and e-
commerce [22]–[25]. The main responsibility is to specify the objectives, diet, and interests of the user.
Recommendation systems [11] are necessary for personalizing the web content, crafting and browsing
experience of user’s typical interests, the tools that are given for communication purposes between the
communications of huge different spaces [26]–[28]. A personalized view consists of ranking the items relevant
to users’ interests. A wide variety of artificial technique (AI) techniques is used including machine-learning
data, user modeling, and parameter satisfaction. Recommendations are necessarily generated for ADA-boost
machine learning algorithms [12]. An Ada-Boost technique is used for the prediction of users' likes and dislikes
in the past [29], the comparative analysis is done based on robust, flexible methods, to train the classifiers. The
results obtained from the Ada-Boost classifier pass through k-fold cross-validation. A novel technique that
established a tourism recommendation system that mines all the user preferences for providing personalized
recommendations [30]. The reviews that are gathered from social media regarding tourism provide a very huge
amount of information for the extraction of preferences. Furthermore, all the comments that are semantically
preprocessed as well as sentimentally analyzed are preprocessed for detecting tourist preferences [31], [32].
Similar to this, the features of these areas of interest are extracted using all the aggregated reviews. We may
create detailed profiles for each user and suggest a method for group suggestions. Instead, of using item
preference profiles as in earlier studies, this is based on the sum of all group numbers for deep profiles.
A hybrid method based on the fuzzy multi-criteria collaborative filtering approach for movie
recommendations [17] takes demographic information and item-based ontological semantic filtering into
account. To identify how each criterion connects to the overall ranking, a neuro-fizzy inference method is
applied. A cosine and Jaccard similarities are generated to assess the general similarity of people or movies,
taking into consideration the effect of co-rated item set cardinality on the validity of similarity measures. A
review of the text-based recommender system. Gathers the data from digital repositories that have been

613
published from the year 2010-2020 collected from the literature. As per the author, this survey majorly displays
the four aspects of text-based recommended system (RS) implemented in the literature survey. The
implemented four aspects are datasets as well as feature extraction techniques and computational approaches
along with the evaluation metrics a human-in-the-loop RS [22], is concerned with urban traffic control with an
agent-based architecture. For this procedure, a regional agent dispatcher is set up to assign operators to carry
out various activities whenever “operation on demand” is required. A daily-dependent operational technique
relating to strategic traffic procedures at the control object level detects these requirements [27], [28].
Is a personality-aware product recommendation system [27] built on the discovery of Meta paths and
user interest mining. Even when the user's history does not contain the things or ones similar to them, the Meta-
Interest anticipates the user's interest in these topics as well as the items related to those interests. A fuzzy tree
structure [30] learning activity technique and a learner profile technique for comprehensively describing the
complex learning events along with the learner profiles. The two methods that are fuzzy category trees as well as
related similarity measures are used to conclude the semantic relations within the learning activities or the
learner’s necessity. To establish an overview of the field recommender system [23] as well as to define the current
form of recommendation technique, they are further classified into three major classes. These classes are content-
based as well as collaborative along with a hybrid recommendation method, having incorporated the contextual
data within the recommendation method and providing the support for the multi-criteria ratings as well as
providing flexible along with less intrusive forms of recommendation. An efficient matrix factorization
recommendation system [24] for protecting the user’s privacy by implementing the local differential privacy
technique. As per this algorithm, the established user's side of the rating data is adjusted to lessen the overall
sensitivity. Before transmitting the sensitive data to the aggregator, Laplace noise is additionally added. Now, the
MF algorithm is used to recognize the rating prediction based on the dispersed data.
A chaotic-based reversible data transform (RDT) technique [33], [34] for privacy-preserving data
mining (PPDM) within the recommendation system. By using this approach, the RDT parameter results will
be created locally, negating the need for earlier sharing of the parameter results concerning the recovery
method. This method can be used as an alternative to the standard RDT algorithm [35], where memory and
bandwidth are critical considerations. E-commerce system [14], [15] provides a huge number of products for
thousands of visitors. This system suffers from many issues personalization problems, privacy problems, and
cold start problems. A semantic recommendation model is used to provide system-based recommendations.
The precision of coverage for generated recommendations achieves higher performance. Semantic technology
encapsulates the preferences, which assist in generating the necessary relationships provided by their different
preferences. The semantic structure achieves high performance for the active node to involve a non-semantic
one. Several limitations regarding the present recommendations techniques can be rectified and help improvise
the recommendation capabilities [31] and result in the widespread use of recommender systems over various
ranges of applications. U2CMS is a sequential recommender system that combines collaborative and content-
based similarity models with Markov chains. Data about contents and sequential patterns are both included in
U2CMS [32] to accurately determine the link between objects. A framework is established for data-driven,
knowledge-driven as well as cognition-driven systems Recommender systems called RS [36] systems for
cognitive recommenders. A cognitive recommender is a new kind of intelligent recommender system that will
comprehend all users' preferences and offer them a better recommendation while getting beyond the drawbacks
of cutting-edge methods [37], [38].
A new method known as a group recommendation model with two-stage deep learning (GRMTDL)
has been developed to address the issues of sparse group-item interaction. This algorithm consists of two
sequential stages: group representation learning (GRL) and group preferences learning (GPL). Here, two
cutting-edge attention techniques for the suggestions system are suggested. The contextual item attention
module gathers contextual information, and as a result [39], [40], the pattern and the items adapt to reflect the
user's preferences. The multi-head attention technique increases the user's preference diversity to accommodate
shifting preferences. Recommendation systems have been used for improvising performance because of the
current approaches in deep learning and the knowledge graph [41]. However, the majority of existing
recommendation systems [16], [17] are one-way. To solve the problems of sparse group-item interaction, a
new technique called a GRMTDL has been created. The GRL and GPL stages of this approach are sequential.
Two state-of-the-art attention strategies for the recommendations [42] system are proposed in this section. The
pattern and the items adjust to match the user's preferences because of contextual information being gathered
by the contextual item attention module. The user's preference diversity is increased by the multi-head attention
technique to account for changing preferences. The well-trained algorithm, once implemented online, can
determine the most appealing things for its customers, resulting in a precisely personalized suggestion [43],
[44]. It is presumed that the user's behavioral features have been adequately reflected from past data at this
point and that they will always remain constant throughout time.

 ISSN: 2252-8938
614
3. BACKGROUND
3.1. Recommendation systems
In each industry, recommendation systems are utilized to give clients priority based on their previous
preferences. There are two recommender problems in the recommendation problem. The ability to predict a
user's preferences for a product or to rank and propose products to a user by creating a Top-N list. Three types
of recommendation systems are recognized. one is systems for collaborative filtering that make suggestions to
users based on their interactions, either overtly (for example, past ratings) or covertly (e.g., user comments).
The foundation of content-based recommender systems is the idea that items are suggested based on how much
they resemble previous user favorites. Figure 2 shows the types of Recommendation systems. These methods
are frequently applied to make up for the shortcomings of one method that the other method corrects.
Figure 2. Types of recommendation systems
3.1.1 Content-based filtering (CF) recommender systems
Content-based filtering is implemented to provide similar content as per the user and their previous
activities or feedback are recorded within the system. Memory-based techniques apply a nearest-neighbor
algorithm to identify comparable entities using a user-item rating matrix. As a result, unknown ratings are
projected using the observed entities' prior ratings.
3.1.2. Collaborative filtering (CF) based recommender systems
The technique of filtering for gathering data or the pattern as per the user preferences incorporates
several processes such as employing strategies as well as collaboration within numerous agents and viewpoints,
and data sources; hence, it is called collaborative filtering. We implement collaborative filtering techniques for
huge datasets [2]. Sensing and monitoring data such as financial services institutions, comprises several
financial resources or e-commerce and web apps, where the focus is on the user’s data.
3.1.3. Hybrid recommendation systems
A unique kind of recommendation system called A hybrid recommendation system incorporates
content filtering and collaboration strategies. Hybrid recommendation systems are made up of many single
recommendation systems. This hybrid approach was created as a solution to the problem with conventional
recommendation systems. Two main topics have been the focus of this field's research: the stability vs plasticity
issue and the cold-start issue. When learning-based strategies like collaborative, content-based, and
demographic recommendation algorithms are applied, the cold-start issue arises.
4. DEEP LEARNING-BASED RECOMMENDATION SYSTEMS
A subset of machine learning known as deep learning (DL) learns from many levels of data abstraction
and representation. To enhance customer experience, some tech companies are already using DL systems based
on different neural networks (NNs). Spotify uses convolutional neural networks, while YouTube, eBay, Yahoo,
and Twitter use deep neural networks (DNNs) (CNNs). DNNs and CNNs are just a couple of the different
kinds of networks that have been employed; the list of deep learning algorithms is endless. Why do we need
different kinds of them? The answer relates to the business domain, a particular task, or a recommender
scenario. Alternative NNs or even a hybrid method, depending on the use case, could.

615
4.1. Convolutional neural networks-based recommender systems
CNNs are a good fit for processing unstructured multimedia data if sufficient feature extraction is
performed. They are utilizing video, audio, text, and image data. Knowledge graphs, protein-interaction
networks, social networks, and other non-Euclidean data (non-ordinal or hierarchical data) can also be
processed using CNNs. This type of technology may, for example, be used to provide Pinterest
recommendations. CNN aids in the elimination of the cold start problem and the enhancement of older methods
such as collaborative filtering. This is an important aspect of e-commerce because most shoppers base their
judgments on the appearance of the goods.
4.2. Recurrent neural networks-based recommender systems
We can implement recurrent neural network (RNN) within the recommendation system as it can play
a major role in processing sequential input by creating temporal dynamics of interactions and sequential
behavior methods. We can explain this with an example of YouTube; here you will get a content
recommendation as per the specific time and that will forecast the next content as per based on the previous
one. Nowadays, it is not mandatory to log in for navigating a particular website. All the websites have a cookie
system (also known as a session mechanism), that will help in this.
5. CHALLENGES ASSOCIATED WITH RECOMMENDATION SYSTEMS
5.1. Cold start problem
This type of problem occurs because whenever any new user is added to the recommendation system,
that new user is unable to be detected by the system as it lacks in rating or reviews. This issue results making
it the user difficult for to forecast the user preferences or this can result in less accurate suggestions. We can
get this by an example, whenever any new movie releases then in that case the particular movie has received
very few ratings or reviews so that movie will get fewer recommendations.
5.2. Sparsity
Whenever you buy something or watch something and do not post the review and rating regarding
that user then it results in a sparsity rating model. This will lead to data sparsity problems within the model.
Further, reduces the likelihood of getting similar user groups of rating and reviews.
5.3. Synonymy & Privacy
Whenever any review or rating is displayed by two or more similar names or if some list of objects
by same meaning, then in that case the recommendation system is unable to differentiate that the term displays
a similar item or a different item. Whenever you post a rating or review, you have to post your personal
information within the recommendation system for getting some extra information or services in the future.
This personal information of any reviewer raises concerns about his/her data privacy.
5.4. Scalability
One of the major significant problems is the scalability of algorithms regarding the real-world datasets
within the recommendation system. In this type of model where we get regular entries in that case, the data is
in huge amounts and it changes regularly. That data will be gathered from user ratings and reviews.
5.5. Latency
The latency problem occurs in the recommendation system because daily, new products are being
added within the system. Here we get recommendations for the previously existing products, as the new
products do not have any review or ratings. We can implement collaborative filtering as well as a category-
based method for user-item interaction to resolve this issue.
6. EVALUATION OF RECOMMENDATION SYSTEMS [31]
Recommender systems are prediction models with algorithms that aim to reduce a function's
inaccuracy as much as possible. As a result, it's critical to assess their prediction inaccuracy by comparing
expected results to those produced by the model. The most prevalent measures are presented here.
Recommendation system is evaluated on four parameters that are Precision & Recall, mean square error (MSE),
root mean square error (RMSE) and MSE. Precision understands how capable the system is to deliver the
applicable features with the least volume of recommendations as shows in (1). The recall is how capable is the
model to find all the relevant elements and recommend them to the user as shown in (2). RMSE can be defined
as penalizing huge errors and the rest of the errors are squared as explained in (3). We can say that RMSE is
the square of MSE as shown in (4). A mean absolute error can be defined as the calculation of errors within
paired observations that expresses the same phenomenon as explained in (5). Here j = Variable; n = number of

 ISSN: 2252-8938
616
data points that are not missing; qi = actual value of time series; 𝒒𝒋
̂ = time series estimated as mentioned in (3)
and (4) and hj = prediction; gj = true value; n= number of data points as mentioned in (5). Table 1 shows the
survey table which shows the previously performed research for recommendation system.
𝑷𝒓𝒆𝒄𝒊𝒔𝒊𝒐𝒏 =
𝑻𝒓𝒖𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆
𝑻𝒓𝒖𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆+𝑭𝒂𝒍𝒔𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆
(1)
𝑹𝒆𝒄𝒂𝒍𝒍 =
𝑻𝒓𝒖𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆
𝑻𝒓𝒖𝒆 𝑷𝒐𝒔𝒊𝒕𝒊𝒗𝒆+𝑭𝒂𝒍𝒔𝒆 𝑵𝒆𝒈𝒂𝒕𝒊𝒗𝒆
(2)
𝑴𝒆𝒂𝒏 𝒔𝒒𝒖𝒂𝒓𝒆𝒅 𝑬𝒓𝒓𝒐𝒓 =
∑ (𝒒𝒋−𝒒𝒋
̂)
𝟐
𝒏
𝒋=𝟏
𝒏
(3)
𝑅𝒐𝒐𝒕 𝒎𝒆𝒂𝒏 𝒔𝒒𝒖𝒂𝒓𝒆 𝑬𝒓𝒓𝒐𝒓 = √∑ (𝒒𝒋−𝒒𝒋
̂)
𝟐
𝒏
𝒋=𝟏
𝒏
(4)
𝑴𝒆𝒂𝒏 𝑨𝒃𝒔𝒐𝒍𝒖𝒕𝒆 𝑬𝒓𝒓𝒐𝒓 =
∑ |𝒉𝒋−𝒈𝒋|
𝒏
𝒋=𝟏
𝒏
(5)
Table 1. Survey table
Title Author Algorithms/Methods/
Technology used
Research gap
Hybrid recommender system
for tourism based on big data
and ai: a conceptual
framework
Khalid AL Fararni, Fouad
Nafis, Badraddine
Aghoutane, Ali Yahyaouy,
Jamal Riffi, and
Abdelouahed Sabri
Hybrid tourism
recommender system
architecture
Once the sets of elements considered
relevant to the tourist are selected, the
system plans an appropriate trip by
combining these items using operational
research technics.
Hierarchy visualization for
group recommender systems
Wei Wang; Guangquan
Zhang Jie Lu
Hierarchy visualization
method for group
recommender (HVGR)
Easily extended to individual rss through the
use of a single-member group. An
implementation has been developed and
feasibility is tested using a real data set.
Multi-criteria review-based
recommender system-the
state of the art
Sumaia Mohammed Al-
Ghuribi and Shahrul
Azman Mohd Noah
Multi-criteria
recommender systems
(MCRSs).
Gain more understanding about the multi-
criteria review-based recommender system
and encourage them to explore the implicit
values of the reviews and utilize them in
future studies.
A hybrid recommender
system for improving
automatic playlist
continuation
Anna Gatzioura; João
Vinagre; Alípio Mário
Jorge; Miquel Sànchez-
Marrè
Hybrid recommender
system for automatic
playlist
To outperform other state-of-the-art
techniques, in terms of accuracy, while
balancing between diversity and coherence.
Profile aggregation-based
group recommender
systems: moving from item
preference profiles to deep
profiles
Le Nguyen Hoai Nam Group recommender
systems,
To address the questions to further refine
deep profile aggregation-based group
recommendations
Context-aware recommender
systems for social networks:
review, challenges, and
opportunities
Areej Bin Suhaim And
Jawad Berri
Context-aware system, Problems such as scalability, novelty, and
trust may be a real challenge in some
application domains necessitating a
substantial overhead on the development of
recommender systems
A survey of recommendation
systems: recommendation
models, techniques, and
application fields
Hyeyoung Ko, Suyeon
Lee, Yoonseo Park, and
Anna Cho
Content-based filtering;
collaborative filtering;
hybrid system;
recommendation algorithm
To expand the research and development of
recommendation systems suitable for the
characteristics of business by application
service field
Creating a recommender
system to support higher
education students in the
subject enrollment decision
A. Jesús F. García, R. Ro-
Echeverría, Juan Carlos
Preciado, José María
Conejero Manzano, and
Fernando Sánchez-
Figueroa
Data mining, decision
support system,
The construction of this decision support
system for students, we intend to increase the
graduation rates and lower the dropout rate.
Profile aggregation-based
group recommender
systems: moving from item
preference profiles to deep
profiles
Le Nguyen Hoai Nam Collaborative filtering,
group recommender
systems
To address these questions to further refine
deep profile aggregation-based group
recommendations.
Causal incremental graph
convolution for
recommender system
retraining
Sihao Ding; Fuli Feng;
Xiangnan He; Yong Liao;
Jun Shi; Yongdong Zhang
Graph convolution
network (GCN)-based
recommender models
---------------------------------

617
7. CONCLUSION
Recommendation systems expand the scope of tailored information retrieval on the Internet. A detailed
survey on recommendation systems of prominent works on deep learning techniques. As a result, recommendation
systems are used for a variety of application sectors that use real-time data from wearable devices and click streams
that often produce better results. As evidenced by a slew of recent articles, recommender systems, and deep learning
have been hot subjects in continued research in recent years. Scholars will get more assistance from this research as
they get an understanding of review-based recommendation systems and this will encourage scholars to discover the
implicit values of reviews and they can further utilize these resources in future studies.
REFERENCES
[1] S. Ding, F. Feng, X. He, Y. Liao, J. Shi, and Y. Zhang, “Causal incremental graph convolution for recommender system retraining,”
IEEE Transactions on Neural Networks and Learning Systems, pp. 1–11, 2022, doi: 10.1109/TNNLS.2022.3156066.
[2] J. Jin, H. Guo, J. Xu, X. Wang, and F.-Y. Wang, “An end-to-end recommendation system for urban traffic controls and management
under a parallel learning framework,” IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 3, pp. 1616–1626, Mar.
2021, doi: 10.1109/TITS.2020.2973736.
[3] A. J. Fernandez-Garcia, R. Rodriguez-Echeverria, J. C. Preciado, J. M. C. Manzano, and F. Sanchez-Figueroa, “Creating a
recommender system to support higher education students in the subject enrollment decision,” IEEE Access, vol. 8, pp. 189069–
189088, 2020, doi: 10.1109/ACCESS.2020.3031572.
[4] A. Bin Suhaim and J. Berri, “Context-aware recommender systems for social networks: review, challenges and opportunities,” IEEE
Access, vol. 9, pp. 57440–57463, 2021, doi: 10.1109/ACCESS.2021.3072165.
[5] D. Ayata, Y. Yaslan, and M. E. Kamasak, “Emotion based music recommendation system using wearable physiological sensors,”
IEEE Transactions on Consumer Electronics, vol. 64, no. 2, pp. 196–203, May 2018, doi: 10.1109/TCE.2018.2844736.
[6] A. Gatzioura, J. Vinagre, A. M. Jorge, and M. Sanchez-Marre, “A hybrid recommender system for improving automatic playlist
continuation,” IEEE Transactions on Knowledge and Data Engineering, pp. 1–1, 2019, doi: 10.1109/TKDE.2019.2952099.
[7] S. M. Al-Ghuribi and S. A. Mohd Noah, “Multi-criteria review-based recommender system-the state of the art,” IEEE Access, vol.
7, pp. 169446–169468, 2019, doi: 10.1109/ACCESS.2019.2954861.
[8] W. Wang, G. Zhang, and J. Lu, “Hierarchy visualization for group recommender systems,” IEEE Transactions on Systems, Man,
and Cybernetics: Systems, vol. 49, no. 6, pp. 1152–1163, Jun. 2019, doi: 10.1109/TSMC.2017.2699690.
[9] K. Al Fararni, F. Nafis, B. Aghoutane, A. Yahyaouy, J. Riffi, and A. Sabri, “Hybrid recommender system for tourism based on big
data and AI: A conceptual framework,” Big Data Mining and Analytics, vol. 4, no. 1, pp. 47–55, Mar. 2021, doi:
10.26599/BDMA.2020.9020015.
[10] A. Beheshti, S. Yakhchi, S. Mousaeirad, S. M. Ghafari, S. R. Goluguri, and M. A. Edrisi, “Towards cognitive recommender
systems,” Algorithms, vol. 13, no. 8, 2020, doi: 10.3390/A13080176.
[11] Z. Abbasi-Moud, H. Vahdat-Nejad, and J. Sadri, “Tourism recommendation system based on semantic clustering and sentiment
analysis,” Expert Systems with Applications, vol. 167, p. 114324, Apr. 2021, doi: 10.1016/j.eswa.2020.114324.
[12] F. Liu and H. J. Lee, “Use of social network information to enhance collaborative filtering performance,” Expert Systems with
Applications, vol. 37, no. 7, pp. 4772–4778, Jul. 2010, doi: 10.1016/j.eswa.2009.12.061.
[13] B. Yang, Y. Lei, J. Liu, and W. Li, “Social collaborative filtering by trust,” IEEE Transactions on Pattern Analysis and Machine
Intelligence, vol. 39, no. 8, pp. 1633–1647, Aug. 2017, doi: 10.1109/TPAMI.2016.2605085.
[14] D. Ma, S. Li, X. Zhang, and H. Wang, “Interactive attention networks for aspect-level sentiment classification,” IJCAI International
Joint Conference on Artificial Intelligence, vol. 0, pp. 4068–4074, 2017, doi: 10.24963/ijcai.2017/568.
[15] S. H. Sreedhara, V. Kumar, and S. Salma, “Efficient big data clustering using Adhoc Fuzzy C means and auto-encoder CNN,” 2023,
pp. 353–368. doi: 10.1007/978-981-19-7402-1_25.
[16] J. K. Tarus, Z. Niu, and A. Yousif, “A hybrid knowledge-based recommender system for e-learning based on ontology and
sequential pattern mining,” Future Generation Computer Systems, vol. 72, pp. 37–48, Jul. 2017, doi: 10.1016/j.future.2017.02.049.
[17] B. Walek and V. Fojtik, “A hybrid recommender system for recommending relevant movies using an expert system,” Expert Systems
with Applications, vol. 158, p. 113452, Nov. 2020, doi: 10.1016/j.eswa.2020.113452.
[18] F. García-Sánchez, R. Colomo-Palacios, and R. Valencia-García, “A social-semantic recommender system for advertisements,”
Information Processing & Management, vol. 57, no. 2, p. 102153, Mar. 2020, doi: 10.1016/j.ipm.2019.102153.
[19] N. Ranjbar Kermany and S. H. Alizadeh, “A hybrid multi-criteria recommender system using ontology and neuro-fuzzy techniques,”
Electronic Commerce Research and Applications, vol. 21, pp. 50–64, Jan. 2017, doi: 10.1016/j.elerap.2016.12.005.
[20] K. Bauman, B. Liu, and A. Tuzhilin, “Aspect based recommendations,” in Proceedings of the 23rd ACM SIGKDD International
Conference on Knowledge Discovery and Data Mining, New York, NY, USA: ACM, Aug. 2017, pp. 717–725. doi:
10.1145/3097983.3098170.
[21] S. Kanwal, S. Nawaz, M. K. Malik, and Z. Nawaz, “A review of text-based recommendation systems,” IEEE Access, vol. 9, pp.
31638–31661, 2021, doi: 10.1109/ACCESS.2021.3059312.
[22] J. Jin et al., “An agent-based traffic recommendation system: revisiting and revising urban traffic management strategies,” IEEE
Transactions on Systems, Man, and Cybernetics: Systems, vol. 52, no. 11, pp. 7289–7301, Nov. 2022, doi:
10.1109/TSMC.2022.3177027.
[23] G. Adomavicius and A. Tuzhilin, “Toward the next generation of recommender systems: a survey of the state-of-the-art and possible
extensions,” IEEE Transactions on Knowledge and Data Engineering, vol. 17, no. 6, pp. 734–749, Jun. 2005, doi:
10.1109/TKDE.2005.99.
[24] X. Zheng, M. Guan, X. Jia, L. Guo, and Y. Luo, “A matrix factorization recommendation system-based local differential privacy
for protecting users’ sensitive data,” IEEE Transactions on Computational Social Systems, vol. 10, no. 3, pp. 1189–1198, Jun. 2023,
doi: 10.1109/TCSS.2022.3170691.
[25] Z. Huang, Y. Liu, C. Zhan, C. Lin, W. Cai, and Y. Chen, “A novel group recommendation model with two-stage deep learning,”
IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 52, no. 9, pp. 5853–5864, Sep. 2022, doi:
10.1109/TSMC.2021.3131349.
[26] J.-H. Chiang, C.-Y. Ma, C.-S. Wang, and P.-Y. Hao, “An adaptive, context-aware, and stacked attention network-based
recommendation system to capture users’ temporal preference,” IEEE Transactions on Knowledge and Data Engineering, vol. 35,
no. 4, pp. 3404–3418, Apr. 2023, doi: 10.1109/TKDE.2022.3140387.

 ISSN: 2252-8938
618
[27] S. Dhelim, H. Ning, N. Aung, R. Huang, and J. Ma, “Personality-aware product recommendation system based on user interests
mining and metapath discovery,” IEEE Transactions on Computational Social Systems, vol. 8, no. 1, pp. 86–98, Feb. 2021, doi:
10.1109/TCSS.2020.3037040.
[28] R. Baral, S. S. Iyengar, X. Zhu, T. Li, and P. Sniatala, “HiRecS: A hierarchical contextual location recommendation system,” IEEE
Transactions on Computational Social Systems, vol. 6, no. 5, pp. 1020–1037, Oct. 2019, doi: 10.1109/TCSS.2019.2938239.
[29] W. Dai, Y. Wang, J. Ma, and Q. Jin, “BTR: A feature-based Bayesian task recommendation scheme for crowdsourcing system,”
IEEE Transactions on Computational Social Systems, vol. 7, no. 3, pp. 780–789, Jun. 2020, doi: 10.1109/TCSS.2020.2986836.
[30] D. Wu, J. Lu, and G. Zhang, “A fuzzy tree matching-based personalized e-learning recommender system,” IEEE Transactions on
Fuzzy Systems, vol. 23, no. 6, pp. 2412–2426, Dec. 2015, doi: 10.1109/TFUZZ.2015.2426201.
[31] W. Wei, C. McElroy, and S. Dey, “Towards on-demand virtual physical therapist: machine learning-based patient action
understanding, assessment and task recommendation,” IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol.
27, no. 9, pp. 1824–1835, Sep. 2019, doi: 10.1109/TNSRE.2019.2934097.
[32] Y. Yang, H.-J. Jang, and B. Kim, “A hybrid recommender system for sequential recommendation: combining similarity models
with Markov chains,” IEEE Access, vol. 8, pp. 190136–190146, 2020, doi: 10.1109/ACCESS.2020.3027380.
[33] S. Beg et al., “Dynamic parameters-based reversible data transform (RDT) algorithm in recommendation system,” IEEE Access,
vol. 9, pp. 110011–110025, 2021, doi: 10.1109/ACCESS.2021.3101150.
[34] T. Li, A. Liu, and C. Huang, “A similarity scenario-based recommendation model with small disturbances for unknown items in
social networks,” IEEE Access, vol. 4, pp. 9251–9272, 2016, doi: 10.1109/ACCESS.2016.2647236.
[35] J. Chen et al., “Meta-path based neighbors for behavioral target generalization in sequential recommendation,” IEEE Transactions
on Network Science and Engineering, vol. 9, no. 3, pp. 1658–1667, May 2022, doi: 10.1109/TNSE.2022.3149328.
[36] Z. Huang, X. Xu, H. Zhu, and M. Zhou, “An efficient group recommendation model with multiattention-based neural networks,”
IEEE Transactions on Neural Networks and Learning Systems, vol. 31, no. 11, pp. 4461–4474, Nov. 2020, doi:
10.1109/TNNLS.2019.2955567.
[37] S. Deng et al., “A recommendation system to facilitate business process modeling,” IEEE Transactions on Cybernetics, vol. 47, no.
6, pp. 1380–1394, Jun. 2017, doi: 10.1109/TCYB.2016.2545688.
[38] J. D. West, I. Wesley-Smith, and C. T. Bergstrom, “A recommendation system based on hierarchical clustering of an article-level
citation network,” IEEE Transactions on Big Data, vol. 2, no. 2, pp. 113–123, Jun. 2016, doi: 10.1109/TBDATA.2016.2541167.
[39] L. N. Hoai Nam, “Profile aggregation-based group recommender systems: moving from item preference profiles to deep profiles,”
IEEE Access, vol. 10, pp. 6218–6245, 2022, doi: 10.1109/ACCESS.2021.3140121.
[40] W. Zeng, J. Qin, and C. Wei, “Neural collaborative autoencoder for recommendation with co-occurrence embedding,” IEEE Access,
vol. 9, pp. 163316–163324, 2021, doi: 10.1109/ACCESS.2021.3133628.
[41] R. Zhou et al., “Self-attention mechanism enhanced user interests modeling for personalized recommendation services in cyber-
physical-social systems,” IEEE Transactions on Network Science and Engineering, vol. 9, no. 6, pp. 3835–3846, Nov. 2022, doi:
10.1109/TNSE.2021.3093782.
[42] J. Schafer, J. Ben, J. Konstan, and J. Riedl, “Recommender systems in e-commerce,” in Proceedings of the 1st ACM conference on
Electronic commerce, 1999.
[43] S. Hwang and E. Park, “Movie recommendation systems using actor-based matrix computations in South Korea,” IEEE
Transactions on Computational Social Systems, vol. 9, no. 5, pp. 1387–1393, Oct. 2022, doi: 10.1109/TCSS.2021.3117885.
[44] M. A. Mahmood, N. El-Bendary, J. Platoš, A. E. Hassanien, and H. A. Hefny, “An intelligent multi-agent recommender system,”
2014, pp. 201–213. doi: 10.1007/978-3-319-01781-5_19.
BIOGRAPHIES OF AUTHORS
Shruthi Nagaraj is a Research Scholar in the school of Computer Science and
engineering at Presidency University, Bengaluru. She has a Bachelor's degree in Computer
Science Engineering from VTU, Master's degree in Software Engineering from Ramaiah
Institute of Technology, VTU. She has two years of experience working in the IT industry in
Communication and Network Security. Her area of interests in research are Recommendation
Systems, Deep Learning and Machine Learning Technologies, Artificial Intelligence and
Intelligent Information Retrieval. She can be contacted at: shru.nag95@gmail.com.
Blessed Prince Palayyan is an Associate Professor with the school of computer
science and engineering at Presidency University. He has served in the education sector for
more than 17 years at college level and among the top university in India and has served with
university abroad. He has received his bachelor’s degree from Manonmaniam Sundaranar
University, Master’s degree from Annamalai University and Doctorate from Anna
University, Chennai. His research interest includes major areas of computer science,
particularly security in computing, Recommendation System, Image Processing, Artificial
Intelligence, Robotics and IoT. He can be contacted at this email:
blessedprince@presidencyuniversity.in

Personalized E-commerce based recommendation systems using deep-learning techniques

More Related Content

Similar to Personalized E-commerce based recommendation systems using deep-learning techniques (20)

More from IAESIJAI (20)

Recently uploaded (20)

Personalized E-commerce based recommendation systems using deep-learning techniques