International Journal of Computer Information Systems and Industrial Management Applications

ISSN 2150-7988 Volume 17 (2025) pp. 143–154

©Cerebration Science Publishing

Article

Machine Learning-Based Music

Classification and Recommendation System
from Spotify

Nilesh Bhaskarrao Bahadure 1, *, Muktinath Vishwakarma 2 , Manish P. Kurhekar 2 ,

Jagdish Chandra Patni 3 and Prasenjeet D. Patil 4
1 Symbiosis Institute of Technology Nagpur Campus, Symbiosis International (Deemed University),
Pune 412115, Maharashtra, India
2 Department of Computer Science and Engineering, Visvesvaraya National Institute of
Technology Nagpur, Nagpur 440010, Maharashtra, India; [email protected] (M.V.);
[email protected] (M.P.K.)
3 CSE, School of Advance Computing, Alliance University, Bengaluru 562106, Karnataka, India;

[email protected]
4 MIT School of Engineering, MIT ADT University, Rajbaug, Loni Kalbhor, Pune 412201, Maharashtra,

India; [email protected]
* Correspondence author: [email protected]

Received date: 4 April 2024; Accepted date: 11 June 2024; Published online: 6 January 2025

Abstract: Music holds significant importance in our daily lives, serving as an earnest element. Amidst the vast
expanse of available information, our objective lies in sieving through data to present users with relevant music or
song content aligning with their interests and objectives. To improve music recommendation accuracy and real-time
recommendation ability, we propose a hybrid music recommendation model based on a Popularity-based Song
Recommender System, a Personalized Song Recommender System, and a Content-based Music Recommendation
System. Our endeavor focuses on constructing a music recommendation system, which operates as a filtering
mechanism predicting user preferences in music based on their inclinations. This approach employs content filtering
techniques leveraging data characteristics to refine recommendations. Our model, tailored to accommodate diverse
instances and datasets, demonstrates robust performance. The devised song recommendation system exhibits
notable efficacy and sustained reliability. Encouraging outcomes, boasting an accuracy of 90.4%, furnish novel
insights, laying the groundwork for future exploration in song recommendation systems.

Keywords: K-NN; machine learning; elbow method; content-based recommendation; popularity-based

recommendation; personalized-based recommendation; generative ai

1. Introduction
Music plays an indispensable role in our lives. There is a chunk of music for all moods and reasoning, and it also
acts as a health therapist. There is a considerable amount of information that uses Big Data to advise or recommend
extra information to consumers. It is typically connected to Machine Learning based algorithms [1–4].
These may be based on numerous streamers, such as the most recent search, search history, and other constituents.
The internet stores our search history, and then the same product (what we search for) is repetitively shown to us; it is
very iterative. Sometimes, even though it is irrelevant, the internet has repeatedly shown us the related information.
To mitigate this irritation and irrelevance in the recommendation system, the users of our system find new songs and
services they may have yet to find on their own, which makes our Music Recommend System quite helpful and easy to
use [5,6]. The proposed system makes searching and then recommendations highly commendable and comfortable.
Our system also provides a social aspect, allowing users to share and discover new music with friends and
colleagues. Our music recommendation system is the perfect solution for music lovers looking to expand their
musical knowledge and discover various artists. With a significant commitment to accuracy, diversity, and
personalization, our project provides an unmatched music discovery experience.

Copyright: ©2025 by the authors

 The need for personalized music recommendations has become paramount in today’s digital era, where music
is abundantly available across various platforms. Content-based music recommendation systems offer a promising
solution to this challenge by leveraging the intrinsic characteristics of songs and user preferences to generate tailored
recommendations.
Traditional recommendation systems rely heavily on collaborative filtering, which analyzes user behavior and
preferences to suggest similar items. While effective, collaborative filtering often faces challenges like the cold-start
problem, where new users or items need more data for accurate recommendations [7,8].
Content-based song/music recommendation systems analyze factors such as genre/grouping, tempo,
instrumentation, and lyrics to understand the musical characteristics of each track. By comparing these attributes
with user preferences, listening, and search history, content-based recommendation systems can suggest songs that
align closely with the user’s flavor [9–11].
One of the critical advantages of content-based recommendation systems is their ability to mitigate the cold-start
problem and offer transparency in recommendations, as users can understand why certain songs are suggested by
relying on song attributes rather than historical user data alone and based on their explicit features.
Despite their effectiveness, content-based recommendation systems face challenges, such as the limited scope of
recommendations based solely on song attributes and the difficulty in accurately capturing nuanced user
preferences [12–16].
In this context, our research aims to explore the potential of content-based music recommendation systems in
providing users with personalized and relevant song suggestions. By investigating the effectiveness of similarity
measures and machine learning algorithms, we seek to enhance the accuracy and usability of content-based
recommendation systems in music streaming and discovery. Through our endeavors, we aim to contribute to the
advancement of music recommendation technology and enrich the listening experiences of users.
The main objectives of our work are as follows:
1. To create a system for recommending songs.
2. To suggest new songs that the user is likely to appreciate.
3. To make users quickly find new songs they like, it will help them explore, saving time and effort.
4. Make recommendations by looking for music with features similar to the listener’s taste.
5. To implement data pre-processing and different algorithms of machine learning.
6. To improve the model’s performance and increase accuracy to some extent.

2. Literature Review
One of the best gifts to the human being is music. We are happy, sad, or neutral; whatever the situation, songs play
a commanding role in our life. Music is a way to relate our life’s journey, so listening to the right music at the right
time will boost our mood and concentration. Nowadays, many search options are available through technological
advancement, even though the songs are automatically recommended depending on our search history. However,
sometimes due to the inaccuracy in the recommendation system, the wrong selection of music will be made. It
should be correctly reflected according to our search data, and the keywords we entered should be validated without
any mismatch and hurdles.
Ziyang Yu et al. [17] investigated the facial micro-expressions recognition (FER) technique. According to the
facial expressions, the algorithm recommends the music. The algorithm has extracted the feature vector, and then the
music will be recommended with the help of the cosine similarity index (CSI). The recognition rate of the proposed
system is 62.1%.
Yunzhe Dong [18] proposed a machine learning-based music recommendation system using content-based and
collaborative filtering techniques. For the new music and to solve the interest drift problem, the real-time
recommendation is resolved using a hybrid music recommendation model that uses personalized measurement and
game theory [19]. The interest drift problem in the online recommendation and the thorny problem in the offline
recommendation are mitigated using a game theory-based recommendation system.
The musical emotion analysis and the performance of emotion prediction are very well integrated by Cao and
Park [20] using LSTM networks and IoT-based techniques. This fusion or combined model gives an average of 0.921
absolute error, 0.534 root mean square error, 0.498 R square of Arousal, and an average of 0.902 absolute error, 0.575
root mean square error, and 0.478 R square of Valence. This study reveals significant implications of music emotions
and provides high-quality music recommendations.

144
 In Chinese music, the musical vogue patterns are estimated, and informed support for advertising and
sponsorships is provided using a Dynamic Incremental Network (DINet) based system [21]. The experimental
analysis outperforms the traditional and baseline models by demonstrating superior performance by enhancing
musical vogue pattern estimation.
In research [22], a combination of machine learning-based algorithms such as Random Forest, Extra Trees,
Gradient Boosting, Multi-Level Perceptron, AdaBoost, Decision tree, Logistic Regression, K-Nearest Neighbors,
Nearest Centroidare, Bernoulli Naive Bayes, Quadratic Discriminant Analysis employed, and Radius Neighbors for
classification and prediction. This study explores audio features on the Spotify dataset and evaluates the feature’s
importance for prediction from the playlist in Indonesia. It achieved the highest accuracy, 69.74%, for the Random
Forest-based technique, followed by the accuracy for Extra trees with 68.07%.
A popular music dataset, GTZAN, is used for analyzing various types of audio signals and music features.
Multiple machine learning algorithms such as naive Bayes, logistic regression, decision trees, and random forests
are supported by Apache Spark and explored for music grouping classification. The experimental analysis shows
that the random forest-based classifier is outperformed with 90% accuracy, and it also works very well for the
mislabelled and distorted dataset [12].
The Internet of Things network and Edge Computing have been investigated for the vocal music signing learning
system [13]. This research applies IoT perception to piano technology and uses edge computing-based algorithms for
deploying sensors into the system and making the system more humanized and intelligent.
An improved music trend prediction and recommendation system based on LSTM and a random forest-based
machine learning algorithm is presented for pop music [9]. The system’s performance is analyzed using the evaluation
indexes root means square error (RMSE) and mean absolute error index (MAER), and the results indicate better music
prediction.
From the enormous song data, selecting the anticipated data is one of the most arduous tasks. Gao Yang [23]
presented a technique that predicted the user behavior, and then an automatic tag generation system recommends the
music. The low coverage problem is also eliminated by combining depth content with other information. This
combined effect improves the tag quality and achieves better classification results of song recommendation
accordingly.
Xinglin Wen [24] has investigated the support vector machine (SVM) based classification technique and the
multi-scale feature extraction using a fast RCNN algorithm for recommending music to the user. The system uses
deep learning and Internet of Things (IoT) based architecture and techniques. A detailed analysis has been done for
recommending music in different indoor scenarios such as a bathroom, bedroom, etc. The use of smartphones is now
very common; almost 80% of users carry smartphones and listen to music on them. The idea of analyzing mobile
accelerometer data and then, based on human activity recognition, the music recommendation system is implemented
by Hyoung-Gook Kim et al. [25]. To access the accelerometer data from the smartphone, authors have utilized a deep
residual bidirectional gated recurrent neural network. The experimental analysis of real data proves the validity of
the proposed system.
Noor Azilah Muda et al. [26] has investigated contents based music recommendation and Modified AIS-based
algorithm for music classification system. Based on the features selection, the songs are classified according to the
genres. The system is capable to analyzes rhythm, timbre, and pitch-based features for the classification of
music genres.
Hiromu Yakura et al. [27] proposed a music recommendation system with matching user preferences while
working. The system considers the user’s concentration level. Then it prioritizes the unplayed songs, estimates the
preference level, and recommends the music. The entire system works based on the Focus Music Recommender
system. Table 1 gives summary of related works.

145
Table 1. Summary of related works.

Authors Technique Used Findings

Aldiyar Niyazov et al. Deep learning and The main idea to develop this recommendation
[28] computer vision system is based on the external features of musical
compositions, such as genre, artist, title, tags,
etc., but on the acoustic similarity of musical
compositions.
Adiyansjah et al. [29] Convolutional Recommendation from the music database has
Recurrent Neural been given based on the similarity of features in
Networks audio signals.
Qiu and Jia [30] Internet of Things & Western music recommendation system is
CNN investigated using the Internet of Things and
CNN-based classifier and achieved an accuracy
rate of 96.7% and a user satisfaction ratio of
98.6%.
Meng Lu et al. [31] Deep learning The proposed system integrates two methods
for interlinking a content-based music
recommendation system, and the entire
methodology of extracting music audio features
has been improved using deep learning techniques.
Chen Chen [32] Deep learning The author has presented a two-layer attention
mechanism by considering music names through
textual CNN and music label text data. The overall
quality and normalized discount cumulative gain
(NDCG) is improved with a high standard
characterization ability.
Rushabh Chheda et Transfer learning and This study deals with a music recommendation
al. [33] CNN from emotion matching through image visuals.
Bhavyajeet Singh et Natural language In this study, Reddit, a social media platform, is
al. [34] processing used to analyze mood in the context of depression
and recommend music accordingly.
Vijay Prakash Neural network This paper talks about how the music
Sharma et al. [35] recommendation system that is being existed for a
long time, in most of them, the recommendation
is decided once the user’s preferences over time.
It also suggests a neutral network-based approach
to song recommendation where their facial
expressions helps detect mood of a person.
Wang and Wang [36] Deep belief network It talks about the two-stage approach of content-
and probabilistic based music recommendation system, where first
graphical model they extract traditional audio content features and
then predict user preferences.
Singh and Dembla Transfer learning Emotions based music system is recommended.
[37] Webcam is used to extract facial traits and is
then recognized using transfer learning techniques.
Spotify songs dataset is used for experimental
analysis.

146
3. Proposed Methodology
Big Companies like Spotify and Apple Music use a hybrid approach to recommendation, applying more than one
method to recommend user songs and artists. Their model includes content-based as well as collaborative filtering
methods. A hybrid recommendation is conventionally used in companies that manage large databases like Amazon.
There are other works done to analyze the accuracy of music recommendation, one of which is using deep learning
to enhance the quality of music recommended to the user and increase accuracy that cannot be otherwise increased
through traditional filtering methods.
Our Music Recommendation System uses unsupervised learning which uses machine learning to analyze and
cluster unlabeled datasets [38]. It uses the K-means clustering model. Using k-means we can group similar data
points into k- distinct clusters. It is based on how far two data points are, the points are said to be similar if the
distance between the two points are closer to each other. Implementation of k-means can be done using centroids
which are the center of the data points. The proposed system has six stages, as shown in Figure 1.

Data Storage

Data
Preprocessing

Feature Extraction
and Selection

Grouping and
Clustering

Spotify Songs Classification

Dataset

Recommendation

Figure 1. Proposed system.

3.1. Data Storage and Pre-Processing

The data used to implement our algorithm is taken from the Kaggle dataset with the file Spotifyfeatures.csv.
This CSV file is extracted from Kaggle. The pre-processing is applied to ensure the dataset data are free from any
noses and other oddities and also to ensure no data or information is missing. Pre-processing is a process to get a
clean dataset from filthy data. The results are also validated without pre-processing, but the system’s accuracy is
compromised. In Python programming, using Pandas data frame and using basic commands such as df.head()
and df.isnull().sum() are used for getting the overview of the data and getting the number of missing values by
reading each column of a data frame.

3.2. Feature Extraction and Selection

We have used a correlation matrix to find the correlation between various features. Suppose the value is less
than 0, i.e., negative. In that case, it indicates that the particular features are inversely related, meaning one feature
increases, the other will decrease. In contrast, if the value is positive, it indicates that the features are directly related,
and if one increases, others will also increase. We need to determine which features are correlated with each other.
They will form clusters of similar features, which will then be used to recommend [39–41].

147
3.2.1. Min-Max Scaler
The MinMaxScaler class from sklearn.preprocessing module is imported, which provides many methods
for normalizing data. The min-max scaler is the method where the data has a scale range of 0 and 1. This scaler is
done by subtracting the minimum value from each value and dividing by the range. It creates a list of datatypes that
will be normalized. It includes the following data types: int16, int32, int64, float16, float32, and float64.
Now, the loop over the columns gives us normalized columns using MinMaxScaller class. The output will be a
new Data Frame called normalization which contains the normalized data. The min-max scaler is the method where
the data has a scale range of 0 and 1. This scaler is done by subtracting the minimum value from each value and
dividing by the range.
Normalized data refers to a dataset in which the values have been adjusted or rescaled to fall within a specific
range or distribution. Normalization is commonly applied to ensure that different variables or features in a dataset
have a similar scale [42].

3.3. Clustering and Grouping

To determine the optimum number of clusters, we employed the elbow method. The elbow method uses the within-
cluster sum of squares (WCSS) that counts the sum of squares of the distance between the data and the centroids of
the clusters. The more the number of clusters, the less the value of the WCSS and the point on the x-axis where there
is a sharp drop in the graph, considered the optimal number of clusters [43–45]. The only problem with the elbow
method we encountered was that it could not give a sharp point of the elbow; in that case, we used another way to
find out the number of clusters, or we considered one of the values of the graph. The optimum value of clusters using
the elbow method is shown in Figure 2.

15
1x10 Elbow Method

3.0

2.5

2.0
WCSS

1.5

1.0

0.5

2 4 6 8 10
Number of Clusters
Figure 2. Elbow method for clustering.

3.4. Classification
This study investigates three techniques for classification based on the popularity-based song recommendation
system, personalized song recommendation system, and content-based music recommendation system. The results
are validated through excessive experimentation and found that the content-based music recommendation system
outperforms the other methods. A content-based music recommendation system is commonly used to solve search
and optimization problems for data with constrained and unconstrained relativity.

3.4.1. Popularity-Based Song Recommender System

We will utilize the popularity_recommender_py class from the recommenders library to implement a popularity-
based recommender system. This approach allows the system to recommend items with universally demonstrated
appeal across the user base without tailoring the recommendations to individual user preferences. The core idea

148
is to recommend universally popular items among all users without personalizing these recommendations based on
individual user preferences. This approach is simple and effective for scenarios where personalized data is sparse or
when the goal is to increase the visibility of the most popular items across the entire user base.

3.4.2. Personalized Song Recommender System

We now create an item similarity based collaborative filtering model that allows us to make personalized
recommendations to each user. For this, we will utilize the item_similarity_recommender _py class from the
recommenders library.
The item similarity-based collaborative filtering model represents a tailored approach to recommendations,
leveraging user-item interactions to identify similarities between items and personalize suggestions for each user.
By analyzing patterns in user behavior, this model discerns which items are often liked, rated similarly, or interacted
with in comparable ways, using these insights to predict a user’s affinity for items they have yet to encounter. The
core principle revolves around the notion that if a user has shown a preference for specific items, they are more
likely to appreciate similar items. This method adapts to individual tastes, delivering bespoke recommendations that
align closely with each user’s unique preferences. It shines in environments rich in user interaction data, allowing
for nuanced understanding and prediction of user interests. This personalized approach enhances user engagement
by suggesting items with a higher likelihood of resonating with the user’s taste, contrasting with broader,
popularity-based strategies by focusing on individual users’ specific interests and behaviors.

3.4.3. Content-Based Music Recommendation

The item_similarity_recommender _py class from the recommenders library also provides a function
get_similar_items(), which gives content-based recommendations.
Content-based recommendation systems utilize item attributes and metadata to personalize suggestions, matching
users with new items similar to those they have previously enjoyed. By analyzing specific features like genre or tags, it
crafts user-specific recommendations based on content similarity. This method offers tailored suggestions that closely
align with individual preferences, leveraging detailed item information to enhance user engagement.
The model using a content-based music recommendation system is validated on the Spotify dataset with the
following parameters:
no. of unique songs in the training set: 8,169
Non-zero values in cooccurence_matrix:1,567
The top ten most popular songs from the training dataset of 8169 songs are also analyzed, shown in Table 2.
The percentage obtained is based on the most frequently to the least frequently access of the songs from the similar
grouping of the songs.

Table 2. Most popular ten songs in the dataset.

Song Play Count Percentage

Bjork-Undo 7,032 0.430419
Dwight Yoakam-You’re The One 6,412 0.392470
Kings Of Leon-Revelry 6,145 0.376127
OneRepublic-Secrets 5,841 0.357520
Charttraxx Karaoke-Fireflies 4,795 0.293495
Cartola-Tive Sim 4,548 0.278377
Kings Of Leon-Use Somebody 3,976 0.243366
Usher featuring will.i.am-OMG 3,947 0.241591
Train-Marry Me 3,578 0.219005
Coldplay-The Scientist 3372 0.206396

149
4. Results and Discussion
Here, We build a KMeans class instance with the desired number of clusters set to 4. We can tell the algorithm
how many clusters to look for using the n_clusters option. The dataset’s normalized form is represented by
normalization. Prior to using K-means clustering, it is typical to normalize the data using a method like Min-Max
scaling or standardization (mean = 0, standard deviation = 1). The fit_predict() method is once more invoked
to cluster the normalized data and determine the cluster assignments.
A dataset is subjected to Min-Max scaling using the class MinMaxScaler from the Sklearn package. We applied
the scaling operating, created a MinMaxScaler object, fit it to the data, and then used the scaler to alter the data to
apply Min-Max scaling to the “features” column.
The recommendation method takes a song name as input and returns a list of songs most similar to the input song.
The algorithm works as follows:
1. The method first finds the row in the numeric_df corresponding to the input song.
2. The method then creates a new DataFrame that contains all of the songs in the numeric_df except for the
input song.
3. The method then calculates the Euclidean distance between the input song and each in the new DataFrame.
4. The method then sorts the new DataFrame by distance and returns the top amount of songs.
5. The Euclidean distance metric is a simple way to measure the similarity between two points in a vector space.
The distance between two points is calculated as the square root of the sum of the squares of the differences
between their corresponding coordinates. In this case, the coordinates of the points are the values of the
different features in the numeric_df.
6. The Spotify_Recommendation class is a simple but effective way to recommend songs to users. The class is
easy to use and can be customized to recommend songs based on different features.
The proposed classification algorithms have been validated through experimentation. The response to the top 10
popular songs for the 1st user, 50th user, and 100th user has been checked. It was noticed that the output for all the
users is the same in the case of the Polularity-Based Song Recommender System and Content-Based Music
Recomendation. However, it is different in all cases for the Personalized Song Recommender System. The
responses using the Polularity-Based Song Recommender System, Personalized Song Recommender System, and
Content-Based Music Recomendation for the top ten recommended songs are shown in Tables 3–5. The score and
the rank are produced based on the most recommended to the least recommended from the similar grouping of
the songs.
The responses from the Personalized Song Recommender System and Content-Based Music Recommendation
System are validated from excessive experimentation, and it is found that for the different users, the output for all the
users is the same. It is also noticed that the score obtained using the Personalized Song Recommender System has
ambiguity. However, there are no such things in Content-Based Music recommendations, so Content-Based Music
recommendations outperform the two.

Table 3. Most popular ten songs using popularity-based song recommender.

Song Score Rank

Björk-Undo 0.7032 1.0
Dwight Yoakam-You’re The One 0.6412 2.0
Kings Of Leon-Revelry 0.6145 3.0
OneRepublic-Secrets 0.5841 4.0
Charttraxx Karaoke-Fireflies 0.4795 5.0
Cartola-Tive Sim 0.4548 6.0
Kings Of Leon-Use Somebody 0.3976 7.0
Usher featuring will.i.am-OMG 0.3947 8.0
Train-Marry Me 0.3578 9.0
Coldplay-The Scientist 0.3372 10.0

150
Table 4. Most popular ten songs using personalized song recommender.

Song Score Rank

Fleet Foxes-Quiet Houses 0.050211 1
Fleet Foxes-Tiger Mountain Peasant Song 0.048733 2
Fleet Foxes-White Winter Hymnal 0.048470 3
Fleet Foxes-Your Protector 0.048036 4
Fleet Foxes-Sun It Rises 0.047604 5
Fleet Foxes-Heard Them Stirring 0.046229 6
Fleet Foxes-Meadowlarks 0.046217 7
Fleet Foxes-Oliver James 0.045491 8
Foo Fighters-No Way Back 0.041123 9
Fleet Foxes-Drops In The River 0.040744 10

Table 5. Most popular ten songs using content-based song recommender.

Song Score Rank

Fleet Foxes-Quiet Houses 0.304233 1
Fleet Foxes-Your Protector 0.286962 2
Fleet Foxes-He Doesn’t Know Why 0.281250 3
Fleet Foxes-Meadowlarks 0.258095 4
Fleet Foxes-Heard Them Stirring 0.234336 5
Fleet Foxes-Tiger Mountain Peasant Song 0.228992 6
Fleet Foxes-Sun It Rises 0.215946 7
Fleet Foxes-White Winter Hymnal 0.188730 8
Fleet Foxes-Ragged Wood 0.175576 9
Pearl Jam-Got Some 0.166667 10

When the keyword ‘Saw You In A Dream’ is entered with cluster size 15, then the 15 songs of similar genres are
recommended from the dataset of Spotify. The result obtained for the recommended songs for ‘Saw You In A Dream’
is shown in Figure 3.

Figure 3. Recommended songs for ‘Saw You In A Dream’.

The different keywords (songs) are hit 500 times, and 452 times, the results are matched. Hence the accuracy
of the proposed system is 90.4%. However, from the 37 mismatched results, total eleven times 60% to 75% of the
recommended songs are correctly shown, and only 26 times, it is less than 60%, which is not recommended. So
effectively, the overall accuracy of the proposed system is 94.8%.

151
5. Conclusions and Future Scope
The music recommendation system has significant potential for future development and enhancement. Our
proposed system shows adoptable accuracy, and the results are highly encouraging; these open new avenues of song
recommendation systems using machine learning and other advanced techniques. Here are some potential future
scopes of the system:
1. Integration with new music platforms: The music recommendation system could be integrated with new and
emerging music platforms, such as social media platforms, online music stores, or streaming services. This
could expand the range of available music data and provide new opportunities for user engagement.
2. Use of advanced machine learning techniques: The recommendation system could be enhanced with more
specialized machine learning techniques, such as deep learning techniques or sentimental analysis.
3. Integration with other recommendation systems: The music recommendation system could be integrated
with other recommendation systems, such as movie recommendation systems or book recommendation
systems, to provide a more comprehensive and personalized recommendation experience for users.
4. Expansion of social interaction features: The social interaction module of the system could be expanded with
new features, such as live chat, group discussions, or virtual events. This could enhance user engagement
and create a more interactive community of music listeners.
Spotify’s dataset is used in this study, with 76,353 unique users and 100,000 unique songs. We have also checked
if any value is missing or not. If any value is missing, we have to delete that row with that missing value. Missing
values or entries will affect the algorithm’s performance. In a future study, instead of deleting the missing value, the
new value generation technique by finding the mean of the nearby values will be considered using the K-NN-based
technique. Through the experimental analysis, a cross-check was done to ensure no empty value in the table.

Author Contributions
Conceptualization, N.B.B. and M.V.; methodology, M.P.K.; software, J.C.P.; validation, P.D.P., and N.B.B.; formal
analysis, N.B.B.; investigation, N.B.B.; resources, M.V.; data curation, M.K.; writing-original draft preparation,
N.B.B.; writing-review and editing, J.C.P.; visualization, M.V.; supervision, P.D.P. All authors have read and agreed
to the published version of the manuscript.
Funding
This research received no external funding.
Conflicts of Interest
Author declares no conflict of interest.
Data Availability Statement
The datasets used and/or analyzed during the current study available from the corresponding author on request.

References
1. Bahadure, N.B., Patil, P.D., Kalbande, S., Pipalatkar, S., Nagar, S., Kuhikar, O: Machine learning based music
recommendation system from spotify. In Proceeding of the 15th World Congress on Nature and Biologically
Inspired Computing (NaBIC 2023), (2023)
2. Karishma, M: Music recommendation system using machine learning. International Journal of Research
Publication and Reviews 5(1), 2024, 20–24
3. Schedl, M: Deep learning in music recommendation systems. Frontiers in Applied Mathematics and Statistics
5, 2019,
4. Sharma, V., Bajaj, A., Abraham, A: Machine learning techniques for electronic health records: Review of
a decade of research. International Journal of Computer Information Systems and Industrial Management
Applications 15, 2023, 595—604
5. Chirasmayee, B.V.S., Sharmila, G., Sahithi, D., Prabhakar, V: Song recommendation system using tf-idf
vectorization and sentimental analysis. International Journal for Research in Applied Science & Engineering
Technology 10(VI), 2022, 2475–2483
6. Fathollahi, M.S., Razzazi, F: Music similarity measurement and recommendation system using convolutional
neural networks. International Journal of Multimedia Information Retrieval 10 2021, 43–53
7. Ferilli, S., Redavid, D., Pierro, D.D., Loop, L: An ontology-driven architecture for intelligent tutoring systems
with an application to learning object recommendation. International Journal of Computer Information Systems
and Industrial Management Applications 14, 2022, 297–312

152
 8. Nirban, V.S., Shukla, T., Purkayastha, P.S., Kotalwar, N., Ahsan, L: A machine learning perspective on fake
news detection: A comparison of leading technqiues. International Journal of Computer Information Systems
and Industrial Management Applications 15, 2023, 59–68
9. Liu, X: Music trend prediction based on improved lstm and random forest algorithm. Journal of Sensors 2022,
(Article ID 6450469), 1–10
10. Shokrzadeh, Z., Feizi-Derakhshi, M.R., Balafar, M.A., Mohasefi, J.B: Graph-based recommendation system
enhanced by community detection. Scientific Programming 2023, (Article ID 5073769) 1–12
11. Yu, K., Lim, K., Kim, P: Deep learning-based business recommendation system in intelligent vehicles. Mobile
Information Systems 2023, (Article ID 3704217) , 1–12
12. Chaudhury, M., Karami, A., Ghazanfar, M.A: Large-scale music genre analysis and classification using machine
learning with apache spark. Electronics, 11(16), (2022)
13. Li, Q., Liu, H., Zhao, X: Iot networks-aided perception vocal music singing learning system and piano teaching
with edge computing. Mobile Information Systems 2023, (Article ID 2074890), (2023) 1–9
14. Song, Y., Jiao, X: A real-time tourism route recommendation system based on multitime scale constraints.
Mobile Information Systems 2023, (Article ID 4586047), 1–10
15. Wang, X., Liu, C: Design of personalized news recommendation system based on an improved user
collaborative filtering algorithm. Mobile Information Systems 2023, (Article ID 9898337), 1–9
16. Wu, H., Yang, T., Li, H., Zhou, Z: Air quality prediction model based on mrmr–rf feature selection and issa–
lstm. Scientific Reports 13(12825), 2023, 1–15
17. Yu, Z., Zhao, M., Wu, Y., Liu, P., Chen, H: Research on automatic music recommendation algorithm based
on facial micro-expression recognition. In Proceedings of the 2020 39th Chinese Control Conference (CCC).
2020, 7257–7263
18. Dong, Y: Music recommendation system based on machine learning. Highlights in Science, Engineering and
Technology 47, 2023, 176–182
19. Yun, W., Jian, L., Yanlong, M: A hybrid music recommendation model based on personalized measurement
and game theory. Chinese Journal of Electronics 32(6), 2023, 1319—1328
20. Cao, Y., Park, J: The analysis of music emotion and visualization fusing long short-term memory networks
under the internet of things. IEEE Access 11, (2023), 141192–141204
21. Hou, P., Zhang, X: Dynamic incremental network for estimating chinese musical vogue patterns of a given
artist over time. IEEE Access 12, 2024, 18028–18040
22. Saragih, H.S: Predicting song popularity based on spotify’s audio features: insights from the indonesian
streaming users. Journal of Management Analytics 10(4), 2023, 693–709
23. Yang, G: Research on music content recognition and recommendation technology based on deep learning.
Security and Communication Networks 2022, (Article ID 7696840), 1–8
24. Wen, X: Using deep learning approach and iot architecture to build the intelligent music recommendation
system. Soft Computing 25(02), 2021, 1–10
25. Kim, H.G., Kim, G.Y., Kim, J.Y: Music recommendation system using human activity recognition from
accelerometer data. IEEE Transactions on Consumer Electronics 65(3), 2019, 349–358
26. Muda, N.A., Choo, Y.H., Ahmad, N: Content-based feature selection for music genre classification.
International Journal of Computer Information Systems and Industrial Management Applications 14, 2022,
001–009
27. Yakura, H., Nakano, T., Goto, M: An automated system recommending background music to listen to while
working. IEEE Transactions on Consumer Electronics 32, 2022, 355–388
28. Niyazov, A., Mikhailova, E., Egorova, O: Content-based music recommendation system. In Proceedings of the
2021 29th Conference of Open Innovations Association (FRUCT). 2021, 274–279
29. Adiyansjah, Gunawan, A., Suhartono, D. Music recommender system based on genre using convolutional
recurrent neural networks. Procedia Computer Science 157, 2019, 99–109
30. Qiu, H., Jia, X: Western music history recommendation system based on internet-of-things data analysis
technology. Mobile Information Systems 2022, (Article ID 8920599), 1–12
31. Lu, M., Pengcheng, D., Yanfeng, S: Digital music recommendation technology for music teaching based on
deep learning. Wireless Communications and Mobile Computing 2022, (Article ID 1013997) 1–8
32. Chen, C: Design of deep learning network model for personalized music emotional recommendation. Security
and Communication Networks 2022, (Article ID 4443277) (2022) 1–8

153
33. Chheda, R., Bohara, D., Shetty, R., Trivedi, S., Karani, R: Music recommendation based on affective image
content analysis. Procedia Computer Science 218 2023 383–392; International Conference on Machine
Learning and Data Engineering.
34. Singh, B., Vaswani, K., Paruchuri, S., Saarikallio, S., Kumaraguru, P., Alluri, V: ”help! i need some music!”:
Analysing music discourse & depression on reddit. PLoS ONE 18(07), 2023, e0287975
35. Sharma, V.P., Gaded, A.S., Chaudhary, D., Kumar, S., Sharma, S: Emotion-based music recommendation
system. In Proceedings of the 2021 9th International Conference on Reliability, Infocom Technologies and
Optimization (Trends and Future Directions) (ICRITO). (2021) 1–5
36. Wang, X., Wang, Y: Improving content-based and hybrid music recommendation using deep learning.
In Proceedings of the 22nd ACM International Conference on Multimedia. MM ’14, New York, NY, USA,
Association for Computing Machinery (2014), 627–636
37. Singh, K., Dembla, P: A study on emotion analysis and music recommendation using transfer learning. Journal
of Computer Science 19(06), 2023, 707–726
38. Bahadure, N.B., Juneja, S., Chandra, C., Mahapatra, P.D., Sathe, S., Verma, S: Kinect sensor based indian sign
language detection with voice extraction. International Journal of Computer Science and Information Security
16(4), 2018, 135–141
39. Sassi, I.B., Yahia, S.B: How does context influence music preferences: a user-based study of the effects of
contextual information on users’ preferred music. Multimedia Systems 27 2021, 143–160
40. Bin, S., Sun, G: Matrix factorization recommendation algorithm based on multiple social relationships.
Mathematical Problems in Engineering 2021, (Article ID 6610645) (2021) 1–8
41. Shao, B., Wang, D., Li, T., Ogihara, M: Music recommendation based on acoustic features and user access
patterns. IEEE Transactions on Audio, Speech, and Language Processing 17(8) 2009, 1602–1611
42. Dhattarwal, A., Ratnoo, S., Bajaj, A., Abraham, A: Ensemble transfer learning for robust human activity
recognition from images. International Journal of Computer Information Systems and Industrial Management
Applications 15, 2023, 250–258
43. Cheng, R., Tang, B: A music recommendation system based on acoustic features and user personalities. In
Trends and Applications in Knowledge Discovery and Data Mining; Cao, H., Li, J., Wang, R., (Eds.); Cham,
Springer International Publishing (2016), 203–213
44. Shi, J: Music recommendation algorithm based on multidimensional time-series model analysis. Complexity
2021, (Article ID 5579086), 1–11
45. Sun, P: Music individualization recommendation system based on big data analysis. Computational Intelligence
and Neuroscience 2022, (Article ID 7646000), 1–11

154