CLUSTERING MODELS FOR MUTUAL FUND RECOMMENDATION

© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 435
CLUSTERING MODELS FOR MUTUAL FUND RECOMMENDATION
Aayush Shah1, Aayushi Joshi2, Dhanvi Sheth3, Miti Shah4, Prof. Pramila M Chawan5
1,2,3,4 B.Tech Student, Dept. of Information Technology, VJTI College, Mumbai, Maharashtra, India
5Associate Professor, Dept. of Computer Engineering and IT, VJTI College, Mumbai, Maharashtra, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The mutual fund industry has expanded
significantly, providing investors with several investment
options. Mutual fund information is necessary for investors to
make prudent investments. Yet, novice investors may find the
financial environment to be complex owing to the abundance
of information. A mutual fund recommendation system based
on machine learning and data analytics overcomes this issue.
We have proposed a clustering models for recommending
mutual funds by analyzing theories regarding mutual fund
investments and returns.
Key Words: Mutual funds, Clustering models, K-means,
DBSCAN, Hierarchal, Agglomerative
1. INTRODUCTION
Mutual fund investing has become an integral component of
portfolio management for investors and financial
institutions. Yet, choosing the best mutual funds to invest in
may be difficult owing to the vast numberofpossibilitiesand
the complexity of the elements that affecttheirperformance.
It is essential to accurately forecast the performance of
mutual funds in order to makeeducatedinvestingselections.
In this paper, we have described clustering models for
recommending mutual fund investments. The suggested
model takes into consideration a number of implicit and
explicit parameters, such as expense ratios, fund manager
experience, past performance, and net asset values, in order
to create investment recommendations that correspond to
an investor's preferences and risk profile. The models such
as K-means, hierarchical clustering, and DBSCAN group
mutual funds based on their comparable traits and
performance. This allows the models to offer suggestions
based not just on the characteristics of individual funds, but
also on the performance and behavior of funds with
comparable characteristics. Using cutting-edge clustering
techniques, our models provides a complete solution for
investors seeking to improve their mutual fundinvestments.
2. PROBLEM
2.1 Problem statement
To propose clustering models for recommending mutual
funds. Today, there is a lack of personalized and accurate
recommendations for investors due to the vast amount of
data and the complex nature of mutual funds. The existing
approaches are limited and may not provide a satisfactory
solution for novice investors. Hence, there is a need for an
efficient and reliable recommendation system that can
consider the individual preferences and risk tolerance of
investors to provide tailored recommendations for mutual
fund investments.
2.2 Problem elaboration
With the rise of online trading platforms, retail investors
now have access to a wider variety of investment options,
but the sheer number of options can be overwhelming.
Additionally, many investors maylack thefinancial expertise
to evaluate the risks and returns of different mutual funds
effectively.
A mutual fund recommendation system could provide
personalizedinvestmentadvicebasedona user'sinvestment
goals, risk tolerance, and other relevant factors. However,
designing an effective system would require addressing
several challenges. One of the primary challenges isbuilding
a model that can accurately predict the performance of
different mutual funds based on historical data. This
requires identifying relevant features that are predictive of
mutual fund returns and developing algorithms that can
effectively learn from this data.
Another challenge is ensuring that the system can provide
personalized recommendations that reflect each user's
unique investment goals and preferences. This requires
developing effective methods for capturinguserpreferences
and incorporating them into the recommendation process.
Finally, it is important to ensure that the system is
transparent and easy to use for novice investors.Thismeans
designing an intuitive user interface that explains the
rationale behind each recommendation and provides users
with the information they need to make informed decisions.
Overall, a mutual fund recommendation system has the
potential to empower novice investors and help them
navigate the complex world of mutual fund investments.
However, designing an effective system requires addressing
several technical and user-facing challenges.
3. DATA
3.1 Data collection
We acquired our data from the Value Research Online
website. It is a well-established website that provides
financial information and analysis to help investors make
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 04 | Apr 2023 www.irjet.net p-ISSN: 2395-0072

informed decisions about their investments. The website
offers a wide range of services, including mutual fund
research. Additionally, the website follows strict editorial
policies to ensure the accuracy and reliability of its content.
The data provided includes critical attributes for equity,
debt, and hybrid mutual fund types, which are financial
vehicles that investors can use to invest in the financial
markets. Equity funds invest in stocks and have higher risk
and return potential, while debt funds invest infixedincome
securities with a fixed rate of return and lower risk
compared to equity funds. Debt funds are further classified
based on the duration of bonds they invest in. Hybrid funds
invest in a mix of equity and debt securities, offering a
balanced mix of risk and return potential.
3.2 Data preprocessing
The data contained a lot of shortcomings that needed to be
dealt with before passing it to the machine learning model.
Data was scattered in separate databases with different
schemas. Several records had null values. Hence, data
integrations along with data cleaning steps had to be
performed. Thus, to make the data more disposable,
following data preprocessing steps had to be applied.
1) Data integration
For separate features, data was extracted in a separate csv
file. These columns were different for 3 kinds of mutual
funds, i.e., Equity, Hybrid and Debt. Hence, we created a
common schema was necessary tounifytheserecordsunder
a common dataset.
2) Feature selection
Based on the relevance of all features, only those features
were selected that may help in predicting the mutual funds.
Data cleaning
a) Dealing with null values:
 Records missing critical features:
There are several records in the dataset where
critical features such as Sharpe Ratio, Standard
Deviation, and Sortino Ratio are missing. It is difficult
to evaluate risk involved without these features.
Hence, records without these featureswerediscarded
completely.
 Records missing a non-critical value:
Such features were filled with the average value
(mean) of the whole column.
b) Dealing with duplicates: Duplicates were deleted.
c) Handling Outliers: Weusedgraphical methodssuchas
box plots and whisker plots to determine the outliers.
3) Feature Extraction
a) To make the data more expressive, we converted a
few categorical columns with only a few values, into
one hot encoded vector. Clusteringalgorithmsusually
use numerical data and raw form of categorical data
might be erroneous. Hence, in order for the clustering
algorithms to work more efficiently and remove any
bias, we converted columns such asfundcategoryand
fund style.
b) A few new features were added to extract valuable
information from the existing columns. For
example, the column called ‘date’ was converted to
‘age_in_months’ by applying appropriate
mathematical functions.
To work with manager_tenure, only primary
manager tenure was extracted from an array of
managers.
4) Exploratory data analysis
This step involved analyzing data and comparing
different gestures with each other. This resulted in a
correlation matrix between all the features. Using this
matrix, features which values extremely correlated to
each other had had to be removed in order to remove
the bias. Hence columns such as NAV_latest,
NAV_previous had a correlation of 1. These columns
were combined to form only 1 column called
NAV_latest.
5) Scaling data
Before passing the data to the next step, the data needs
to be normalized or scaled so that bigger values don’t
skew the clustering output. All the numerical values
were scaled
By implementing these steps, wecanensurethatthedataset
is cleaned, filtered, and transformed into a more useful
format for recommendation modeling.
Finally, after performing all these preprocessing steps, the
data contained attributes denotingfundtypelikeequitydebt
or hybrid, fund performance metrics like expense ratio,
returns and fund manager tenure, fund style like growth,
value or blend and several other numerical attributes like
risk factor, net asset value, standard deviation, Sharpe ratio
and standard deviation.

4. CLUSTERING MODELS
We propose four clustering models:
1) K-means: It is used to cluster and partition data into
groups based on similarities by minimizing the sum of
squared distances between centroids and data points.
2) Hierarchical: It is used to group data into clusters in a
hierarchical manner, based on the distance between
data points, without needing to specify the number of
clusters beforehand.
3) Agglomerative: It is a hierarchical clustering algorithm
that starts with each point as a single cluster and
gradually merges them into larger clusters with more
points based on their similarity, until all points belong
to a single cluster.
4) DBSCAN: It is a density-based clusteringalgorithmthat
groups data points togetherthatarecloselypackedand
separated from other clusters, based on a user-defined
minimum number of points and a maximum distance
between them.
The process of analyzing and clustering data involves
various techniques that can assistinidentifyingpatternsand
structures within the data. One such technique is scaling the
data to normalize and standardize it to ensure that different
features or variables are comparable and easier tointerpret.
Scaled dataset was used to implement these four types of
clustering algorithms, i.e. Agglomerative, DBSCAN,
Hierarchy, and K-means. The effectiveness of the different
clusters formed using these algorithms was evaluated and
checked against two metrics, which were inertia and
silhouette.
Inertia measures the sum ofsquareddistancesbetweeneach
point and its assigned centroid in the cluster. A lower inertia
value indicates that the clusters are more tightly packedand
well-separated, which is a desirable outcome. Silhouette
score measures how well each data point fits into its
assigned cluster, by comparing the distance between the
point and other points in its own cluster (cohesion) to the
distance between the point and points in the nearest
neighboring cluster (separation).
A high silhouette score (closer to 1) indicateswell-separated
clusters, while a low score (closer to -1) indicates poorly
separated clusters.
By comparing the results oftheclusteringalgorithmsagainst
these metrics, it was determined which algorithm produced
the most optimal and accurate clusters.
We defined hyperparameter search dictionaries for these
clustering algorithms. The parameters for each algorithm
was specified with ranges of possible values. Additionally, a
dictionary containing a list of features was created to use in
the grid search.
5. OUTPUT
For each combination of model and hyperparameters,
clustering has been performed and the results are recorded.
We compare these models on the basis of the silhouette
score. Fig. 1 shows the top K-means silhouette scores with
maximum score of 0.256 forming 2 clusters havingcountsof
598 and 328. Similarly Fig. 2, Fig. 3 and Fig. 4 shows the top
scores for Hierarchical, Agglomerative and DBSCAN
clustering models respectively along with their cluster
counts.
Fig 1: Top K-means Silhouette score
Fig 2: Top Hierarchical Silhouette score

Fig 2: Top Agglomerative Silhouette score
Fig 4: Top DBSCAN Silhouette score
Average Maximum
Agglomerative 0.209734 0.209734
K-means 0.246271 0.256181
Hierarchy 0.528714 0.591561
DBSCAN 0.743419 0.743419
Table 1: Silhouette scores across different algorithms
5.1 Critical clustering features
In order to determine which aspects of the clustering
approach were the most important, we constructed a
Random Forest Classifier model.
We used hyperparameters like Gini index and entropy to
identify the key features that drive the formation of distinct
clusters in a clustering algorithm.
Fig. 5 shows that the most effective feature while using
agglomerative clustering is ‘Equity_fund_style_Growth’
followed by ‘Standard_Deviation’ and ‘Category_Equity’.
Likewise, Fig. 6, Fig. 7 and Fig. 8 show the most effective
features in the Hierarchical, K-means and DBSCAN methods
respectively. It is clear from the observations that ‘Category’
columns play a major role in almost all the clustering
algorithms to divide the mutual funds into clusters.
Fig 5: Most effective parameters of Agglomerative

Fig 6: Most effective parameters of Hierarchy
Fig 7: Most effective parameters of K-means
Fig 8: Most effective parameters of DBSCAN
6. CONCLUSION
In this study, we successfully implemented various
clustering algorithms, including k-means, DBSCAN,
Hierarchical,andAgglomerative,to effectively clustermutual
funds. We evaluated the performance of these algorithms
using parameters such as Silhouette score and Inertia,
allowing for a comprehensive comparative analysis to
identify the optimal method for clustering mutual funds.
Additionally, we employed the Random Forest algorithm to
determine the most influential features that contributed to
the clustering results. This insightful analysis revealed the
order of importance of the features in the mutual fund
clustering process, providing valuable insights for future
research and investment decision-making.
7. FUTURE SCOPE
Developing an efficient clustering model to analyze and
categorize users into distinct clusters based on the
similarities found in their data points will be the next step.
The suggested methods must be further analyzed to
determine which amongst them gives the best result on the
given dataset. The best clustering algorithm can effectively
group users together based on shared features or
characteristics within a given feature space. Once users are
assigned to their respective clusters, a personalized and
effective recommendation can be generated based on the
cluster to which the user belongs. Importantly, this
recommendation is tailored while taking into careful
consideration the unique constraints and limitations that
apply to each user, ensuring that it aligns with their

preferences, requirements, and other relevant factors. This
approach ensures that the recommendations provided are
highly relevant and valuable,providinguserswith a superior
experience while accommodating their specific needs and
constraints.
Furthermore, a sophisticated recommendation system can
be built that takes into account individual investor
characteristics such as investment horizon, risk profile,
investment type, minimum investment and so on to
recommend the best possible mutual fund schemes to that
particular investor that can aidnoviceaswell asexperienced
investors in choosing the best scheme to invest in out of the
thousands available today.
REFERENCES
[1] Aayush Shah, Aayushi Joshi, Dhanvi Sheth, Miti Shah,
Prof, Pramila M Chawan, “Mutual fund recommendation
system with personalized explanations”, published in
International Research Journal of Engineering and
Technology Volume 9 Issue 11, November 2022
[2] Pei-Ying Hsu, Chiao-Ting Chen, Chin Chou & Szu-Hao
Huang, “Explainable mutual fund recommendation system
developed based on knowledge graph embeddings”,
published in Applied Intelligence Volume 52 Issue 9 on 1st
July 2022
[3] Li Zhanga, Han Zhanga, SuMin Hao, “An equity fund
recommendation system by combing transfer learning and
the utility function of the prospect theory”, published in the
Journal of finance and data science on Volume 4, Issue 4,
December 2018
[4] Chae-eun Par, Dong-seok Lee, Sung-hyun Nam, Soon-kak
Kwon, “Implementation of Fund Recommendation System
Using Machine Learning” published in Journal ofmultimedia
information system, Sept 30, 2021
[5] Prem Sankar Ca, R. Vidyarajb,K.SatheeshKumarb,“Trust
based stock recommendation system - a social network
analysis approach”,publishedinInternational Conferenceon
Information and Communication Technologies -ICICT 2014
[6] Nusrat Rouf , Majid Bashir Malik , Tasleem Arif , Sparsh
Sharma , Saurabh Singh , Satyabrata Aich and Hee-Cheol Ki,
“Stock Market Prediction Using Machine Learning
Techniques: A Decade Survey on Methodologies, Recent
Developments, and Future Directions “ published in MDPI,
Nov 8, 2021
[7] Nghia Chu, Binh Dao , Nga Pham, Huy Nguyen, Hien Tran
“Predicting Performances of Mutual Funds using Deep
Learning and Ensemble Techniques “ published in arXiv.org
School of Statistical Finance, Cornell Universityarchive,Sept
18, 2022
[8] K. Pendaraki, Grigorios Beligiannis, A. Lappa, “Mutual
fund prediction models using artificial neural networks and
genetic programming”
[9] Krist Papadopoulos “Predicting Mutual Fund
Redemptions with Collaborative Filtering”
[10] Yi-ChingChoua, Chiao-TingChen, Szu, HaoHuang,
“Modeling behavior sequence for personalized fund
recommendation with graphical deepcollaborativefiltering”
published in Expert Systems with Applications Volume 192,
April 15, 2022
[11] Giridhar Maji, Debomita Mondal, NilanjanDey,Narayan
C. Debnath, Soumya Sen, “Stock prediction and mutual fund
portfolio management using curve fitting techniques”
published in Journal of Ambient IntelligenceandHumanized
Computing, Jan 2, 2021
BIOGRAPHIES
Aayush N Shah, B. Tech Student, Dept. of Computer
Engineering and IT, VJTI College, Mumbai, Maharashtra,
India.
Aayushi Joshi, B. Tech Student, Dept. of Computer
India.
Dhanvi Sheth, B. Tech Student, Dept. of Computer
India.
Miti Shah, B. Tech Student, Dept. of Computer Engineering
and IT, VJTI College, Mumbai, Maharashtra, India.
Prof. Pramila M. Chawan, is working as an Associate
Professor in the Computer Engineering Department of VJTI,
Mumbai. She has done her B.E.(Computer Engineering) and
M.E.(Computer Engineering) from VJTI College of
Engineering, Mumbai University. She has 30 years of
teaching experience and has guided 85+ M. Tech. projects
and 130+ B. Tech. projects. She has published 148 papers in
the International Journals, 20 papers in the
National/International Conferences/ Symposiums. She has
worked as an Organizing Committee member for 25
International Conferences and 5 AICTE/MHRD sponsored
Workshops/STTPs/FDPs. She has participated in 17
National/International Conferences. Worked as Consulting
Editor on – JEECER, JETR, JETMS, Technology Today,
JAM&AER Engg. Today, The Tech. World Editor – Journals of
ADR Reviewer -IJEF, Inderscience. She has worked as NBA
Coordinator of the Computer Engineering Department of
VJTI for 5 years. She had written a proposal underTEQIP-I in
June 2004 for ‘Creating Central Computing Facility at VJTI’.
Rs. Eight Crore were sanctioned by the World Bank under
TEQIP-I on this proposal. Central Computing Facilitywasset
up at VJTI through this fund which has played a key role in

improving the teaching learningprocessatVJTI.Awarded by
SIESRP with Innovative & Dedicated Educationalist Award
Specialization: Computer Engineering & I.T. in 2020 AD
Scientific Index Ranking (World Scientist and University
Ranking 2022) – 2nd Rank- Best Scientist, VJTI Computer
Science domain 1138th Rank- Best Scientist, Computer
Science, India.

CLUSTERING MODELS FOR MUTUAL FUND RECOMMENDATION

More Related Content

Similar to CLUSTERING MODELS FOR MUTUAL FUND RECOMMENDATION (20)

More from IRJET Journal (20)

Recently uploaded (20)

CLUSTERING MODELS FOR MUTUAL FUND RECOMMENDATION