QFLS: A Cloud-Based Framework for Supporting Big Healthcare Data

Management and Analytics from Big Data Lakes: Deﬁnitions,

Requirements, Models and Techniques

Alfredo Cuzzocrea

1,2 a

and Selim Soufargi

1 b

iDEA Lab, University of Calabria, Rende, Italy

Department of Computer Science, University of Paris City, Paris, France

Keywords:

Big Healthcare Data, Big Data Management, Big Data Analytics, Cloud-Based Frameworks.

Abstract:

This paper introduces deﬁnitions, requirements, models and techniques of QUALITOP Federated Big Data

Analytics Learning System (QFLS), a Cloud-based framework for supporting big healthcare data management

and analytics from big data lakes. QFLS anatomy and main functionalities are described, along with the main

software solutions proposed with the framework.

1 INTRODUCTION

Nowadays, big data management and analytics play

a disruptive role in modern database research (e.g.,

(Chen et al., 2013; Zhang et al., 2015; Chaudhuri,

2012; Cuzzocrea et al., 2014; Campan et al., 2017;

Balbin et al., 2020)), with a special touch on both

theory and systems. When applied to healthcare do-

mains (e.g., (Dash et al., 2019; Sun and Reddy, 2013;

Patil and Seshadri, 2014)), relevant challenges arise,

starting from big data representation to big data in-

dexing, from big data understanding to big data an-

alytics, and so forth. These open issues have been

recently highlighted by authoritative proposals in the

scientiﬁc ﬁeld (e.g., (Ghayvat et al., 2022; Ding et al.,

2022; Parimanam et al., 2022)).

Indeed, healthcare domains dictate special re-

quirements, among that the need for supporting ad-

vanced analytics while ensuring the privacy of in-

put big healthcare datasets is a ﬁrst-class point (e.g.,

(Onesimu et al., 2022; Abbasi and Mohammadi,

2022; Singh et al., 2022)). This is one of the main

goal of the EU H2020 project QUALITOP (QUALI-

TOP, 2023). QUALITOP aims at devising a collec-

tion of models, techniques and algorithms for sup-

porting prediction and recommendation for the qual-

ity of life of cancer patients treated with the modern

immunotherapy treatment (e.g., (Vinke et al., 2023;

Beaulieu et al., 2022)).

https://orcid.org/0000-0002-7104-6415

https://orcid.org/0009-0000-5476-9403

One of the most distinctive characteristics of

QUALITOP is represented by the so-called QUALI-

TOP big data lake, a big data lake (e.g., (Cuzzocrea,

2021)) oriented to collect anonymized data/informa-

tion from a reference data federation built at the med-

ical premises, and to support big data analytics and

big data predictive analytics over such datasets. This,

with the ﬁnal goal of supporting analysis and recom-

mendations of immunotherapy-treated cancer patients

during their post-trauma life. The big data lake re-

lies within the general QUALITOP software platform

(e.g., (Elgammal and Kr

amer, 2021)).

In order to fulﬁll the requirements super-imposed

by such scenarios, we propose the QUALITOP Feder-

ated Big Data Analytics Learning System (QFLS), a

Cloud-based federated framework for supporting big

healthcare data management and analytics from big

data lakes. Among various results, QFLS introduces

the so-called Tree-Like Analytical Query (TLAQ)

model, a powerful analytical model for healthcare

analytics. Given a TLAQ Q, every node n ∈ Q is

equipped by query node Q

, modeled as follows:

= ⟨A

, Σ

, P

⟩, such that: (i) A

is the target query

attribute; (ii) Σ

is a constraint over A

; (iii) P

is an

aggregate operator over A

. Another relevant charac-

teristic of the TLQA model is that, given two query

nodes Q

and Q

, such that Q

≺ Q

, there not ex-

ists a strict hierarchical relation between Q

and Q

meaning that A

̸⊂ A

. This special feature is particu-

larly suitable to support precision medicine processes.

When executed, TLAQ generate a tree-like analytical

422

Cuzzocrea, A. and Soufargi, S.

QFLS: A Cloud-Based Framework for Supporting Big Healthcare Data Management and Analytics from Big Data Lakes: Deﬁnitions, Requirements, Models and Techniques.

DOI: 10.5220/0012092700003541

In Proceedings of the 12th International Conference on Data Science, Technology and Applications (DATA 2023), pages 422-428

ISBN: 978-989-758-664-4; ISSN: 2184-285X

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

data structure that indexes anonymized datasets, made

available for several privacy-preserving big data ana-

lytics purposes (e.g., (Lin et al., 2016)).

In more details, QFLS is a data federation solu-

tion that is based on the big data processing frame-

work Hadoop cluster, to which all queries are dis-

patched, In addition, specialized Apache-Spark-based

servers that run remotely within each of the data fed-

eration sources are introduced. QFLS is developed

and deployed on top of the iDEA Lab Cloud, located

at University of Calabria, Rende, Italy, which is com-

posed of 21 VMs equipped with Windows Server Es-

sentials 2019 OS, each having 32GB of memory, 8-

Core CPUs, 32GB RAM and 60GB HDD.

It goes without saying that it is of best interest to

run Spark jobs on the Hadoop cluster in a MapRe-

duce manner, especially for larger datasets, and to

be able to do that, ﬁne tuning Spark job submission

onto the cluster was necessary. Using full potential

of the Hadoop cluster is a key to bring effectiveness

and efﬁciency during query executions. Speciﬁcally,

the server runs on Hadoop (through YARN resource

manager) with ﬁne-tuned worker instances, worker

memory and CPU, and driver memory and CPU (e.g.,

(Chen et al., 2017; Gounaris and Torres, 2018)).

The latter sub-system represents the QFLS Core

component of QFLS, which is the Cloud-based en-

gine that runs everything. QFLS encompasses more

two components, namely QFLS-ADPT and QFLS-

ADAT. The QUALITOP Anonymized Dataset Popula-

tion Tool (QFLS-ADPT) is a web-based tool for man-

aging the anonymized dataset at the federation node.

The QUALITOP Anonymized Dataset Analytics Tool

(QFLS-ADAT) is a web-based tool for supporting big

healthcare analytics and predictive analytics over the

anonymized datasets stored in the federation, via the

TLAQs.

2 QFLS ANATOMY

Figure 1 shows the main blueprint of the big data pro-

cessing ﬂow supported by QFLS. Here, we introduce

an example scenario where the following two nodes

occur: one QFLS federated node, located in France,

and one QFLS Core node, located in Italy. There-

fore, the French node provides (anonymized) health-

care data, and the Italian node support big data ana-

lytics and predictive analytics over these data.

As shown in Figure 1, at the French node (1),

where the target healtcare dataset D is located, med-

ical operators generate an anonymized version of

D, denoted by D

′

(3), according to speciﬁc med-

ical guidelines (2), yet compliant with the GDPR.

Then, the analytics tools located at the Italian node

(4) execute aggregate queries, deﬁned by the input

TLAQ (8), which, in turn, is driven by target big

data analytics tools deﬁned within the QUALITOP

big data lake, over the remote French node, via feder-

ated query algorithms based on Apache Spark, and

the anonymized representation of the result is so-

obtained (5). The ﬁnal TLAQ analytics answer is

shaped as a tree (9) such that each node indexed a

proper anonymized healthcare dataset, empowered by

the Cloud computing potentialities (e.g., distribution,

indexing, load balancing, mirroring, and so forth).

The latter data structure is accessed by medical de-

cision makers (10) who provide the ﬁnal big data an-

alytics and predictive analytics (11).

It should be noted some relevant characteristics of

QFLS, which we summarize as follows:

1. Data Federation. While QFLS Core is entirely

Cloud-based, the main processing of QFLS hap-

pens over data federation nodes, according to

severe data privacy protection guidelines. This

scenario fully converges with real-life systems,

where data federation participants are interested

in participating to the federation (for data analyt-

ics purposes) but are not willing to unveil their

data. The latter is the common case of medical

centers.

2. Privacy-Preserving Advanced Analytical Mod-

els. QFLS encompasses the TLAQ analytical

model, a powerful privacy-preserving analyti-

cal model particularly target to support preci-

sion medicine processes that, by deﬁnition, are

built upon lazy aggregations. For instance, at

ﬁrst, physicians may be interested in analyzing

COVID-19 data about female patients who live in

Canada and whose age in between the range [25-

50], then, within this range, they may want to an-

alyze COVID-19 data about female people of age

30 who are resident in Toronto ans have also been

diagnosed with Tuberculosis.

3. Big Data Analytics & Big Data Predictive Ana-

lytics. QFLS not only supports big data analytics,

mostly driven by the TLAQ model, but also the

so-called big data predictive analytics. Indeed, re-

trieved TLAQ analytics can be combined in a mul-

tidimensional fashion, and so-derived summary

data can be used to develop fortunate big multidi-

mensional analytics metaphors (e.g., (Cuzzocrea

et al., 2004)).

4. Flexible Big Data Processing Tools Integration.

QFLS fully relies on the smoothly integration of

several big data processing tools, mostly within

the eco-system deﬁned by Hadoop, such Apache

QFLS: A Cloud-Based Framework for Supporting Big Healthcare Data Management and Analytics from Big Data Lakes: Deﬁnitions,

Requirements, Models and Techniques

423

Figure 1: QFLS Main Blueprint.

Spark, YARN, Hive, MongoDB, etc. This en-

sures high data-availability, high scalability and,

moreover, a full tendency to be further integrated

within more complex big data stack architectures.

3 RELATED WORK

The problem of supporting big healthcare data man-

agement and analytics has been of great interest re-

cently, even due to the COVID-19 outbreak (e.g., (Liu

et al., 2023b; Dimitsaki et al., 2023)). In this Section,

we provide a brief overview of most relevant propos-

als in this scientiﬁc area.

(Ghayvat et al., 2022) focuses on Internet of

Things (IoT)-based Healthcare services, which are

becoming more widespread today, continuously gen-

erate huge amounts of big data. Due to the data mag-

nitude, data intricacy, privacy preservation, data in-

tegrity and identity veriﬁcation requirements, indeed

novel research challenges and issues in healthcare big

data service management arise. To overcome these

problems, author propose a scalable computing sys-

tem that provides veriﬁable data access mechanism

for IoT-enabled health data analytics in the big data

ecosystem. There are two primary sub-architectures

in the proposed architecture, namely a big data an-

alytics tracking system and a derived blockchain-

based data storage/access system. This approach

leverages big data systems and blockchain architec-

ture to analyze, and securely store data from IoT-

enabled devices and allow veriﬁed access to the stored

data. The zero-knowledge protocol is used to ensure

that no information is accessible to unauthenticated

users alongside avoiding data linkability. The results

demonstrate the effectiveness of the proposed method

to solve the problems of big data analytics and privacy

issues in healthcare.

(Ding et al., 2022) proposes two differentially

private algorithms, i.e., Output Perturbation with

aGM (OPERA) and Gradient Perturbation with aGM

(GRPUA) for empirical risk minimization, a useful

method to obtain a globally optimal classiﬁer, by

leveraging the analytic Gaussian mechanism (aGM)

to achieve privacy preservation of sensitive medical

data in a healthcare system. Authors theoretically an-

alyze and prove utility upper bounds of proposed al-

gorithms and compare them with prior algorithms in

the literature. The analyses show that in the high pri-

vacy regime, the proposed algorithms can achieve a

tighter utility bound for both settings: strongly convex

and non-strongly convex loss functions. Besides, the

proposed private algorithms are evaluated against ﬁve

benchmark datasets. The simulation results demon-

strate that these approaches can achieve higher accu-

racy and lower objective values compared with exist-

ing ones in all three datasets while providing differ-

ential privacy guarantees.

(Parimanam et al., 2022) notices that, current

health information systems, when coupled big data

trends, fail to maintain a highly organized analysis

and processing of health data for analytics purposes.

In addition, it collects issues that play an impor-

tant role in determining quality. In this article, au-

thors propose a Hybrid Optimization based Learn-

ing technique for Multi-Disease analytics (HOL-

MD) from healthcare big data using optimal pre-

processing, clustering, and classiﬁer. First, authors

introduce a capuchin search based optimization al-

gorithm for pre-processing which removes the un-

wanted artifacts to enhance the detection accuracy.

Second, a modiﬁed Harris Hawks Optimization based

DATA 2023 - 12th International Conference on Data Science, Technology and Applications

424

Clustering (MHHOC) technique is introduced used

to select optimal features among multiple features

which discovers the subgroups and reduce the dimen-

sionality issues. Performance of proposed HOL-MD

technique is evaluated using standard US healthcare-

organization SUSY and HIGGS datasets, and it turns

that existing state-of-art techniques are outperformed

in terms of accuracy, precision, recall and F-measure,

respectively.

4 QFLS: REQUIREMENTS AND

MAIN FUNCTIONALITIES

This Section highlights, mostly from a software-

engineering and component-architecture point-of-

view, the requirements and the main functionalities

of QFLS.

4.1 Data Ingestion and Storage: The

QADPT Component

In order to enable data ingestion into QFLS, feder-

ated nodes are enabled with a client web component

that ensures data can be referred by QFLS, and can be

processed for later analytics, according to FAIR prin-

ciples (Findable, Accessible, Interoperable and Re-

peatable).

In order to ingest the (anonymized) data, current

QADPT provides an interface to select data from lo-

cal ﬁle system as CSV ﬁles. The data are then stored

into Hive table and QADPT notiﬁes QFLS of the

ingested data through updating the data setting of

QFLS federated nodes. After updating the QFLS sys-

tem conﬁguration, users are allowed to see the added

datasets in the respective node and use them for sub-

mitting their TLAQ (federated query answering). To

note that, at ingestion time, data are cleansed and

transformed to meet QFLS standards and constraints

(date format, column values encoding issues, and so

forth). On the other hand, it should be noted that the

QFLS-based anonymization mechanism ensures that

users cannot never access the original data, but only

privacy-preserving summarized versions of them.

QADPT is also capable or removing a dataset

from QFLS federated node or also list existing ones in

a intuitive browsable interface. QADPT implements

a role-based access control, and allows authentication

of medical operator users solely.

QADPT is a client web component implemented

using Spring v. 5.3 and deployed over a Tomcat v.

8.5 instance. Main functionalities of QADPT is to

load ingested data into Hive. The insertion into Hive

is performed through a JDBC client which takes care

of loading the dataset (then, a CSV ﬁle) into a cor-

responding table in Hive, it also, as mentioned, up-

sert the associated (dataset) information into a HDFS-

based conﬁguration system ﬁle (.ini) depending on

whether the node is previously existent in QFLS (up-

date) or the node is newly created (insert).

4.2 Data Analytics: The QADAT

Component

In order to enable data scientists and medical oper-

ators with insightful recommendation on the health-

care data, QFLS provides QADAT, a client web com-

ponent that offers numerous tools to be used, all in

a matter of a deep analysis of data and to eventually

provide recommendations, or, better, predictive ana-

lytics, upon these global analysis. Figure 2 shows the

main workﬂow for predictive analytics supported by

QFLS.

Figure 2: QFLS Workﬂow for Predictive Analytics.

In more details, Figure 2 shows the interaction be-

tween the Cloud-based QFLS framework QFLS Core

and the federated nodes. At each federated node, data

are processed as partial analytical tasks, via the TLAQ

queries, and (partial) results are sent back to the Cloud

for aggregation with other partial results similarly ob-

tained from other nodes. The partial results are then

processed and structured as needed and global analyt-

ics are then derived. QFLS produces from the TLAQ-

shaped global analytics a visual dashboard consisting

of predictive analytics tools. These can be in many

forms: clustering, ROC areas, confusion matrices,

and so forth. These predictive analytics tools serve

for recommendation purposes by the competent med-

ical operators for the patients. It is also worth notic-

ing that the aforementioned QFLS system set-up pre-

serves data locality which would increase anonymity

and produce more tailored analytics. All of the data

QFLS: A Cloud-Based Framework for Supporting Big Healthcare Data Management and Analytics from Big Data Lakes: Deﬁnitions,

Requirements, Models and Techniques

425

processing logic is executed on the QFLS Core, in a

MapReduce fashion.

QADAT is a client web component implemented

using Spring v. 5.3 from which users can analyze

the data stored in all linked federated nodes, in an

anonymized manner. Like for the case of QADPT,

the choice of Spring was made because it is compliant

with the employed JDK version 8, and, with our set

of technological solutions such as those used for the

server (e.g., Spark) and for deployment (e.g., Tom-

cat), that was the most conforming solution among

all the available ones. JWT authentication (Spring se-

curity) along with role based access were also part

of our Spring application. Indeed, data analyst role

has full components access whereas medical opera-

tor modules access is restricted. For instance, a med-

ical operator user cannot use some SQL-based mod-

ules. This general idea could be further extended by

adopting adaptive metaphors developed in the context

of web information systems (e.g., (Cannataro et al.,

2001; Cannataro et al., 2002)).

In order to test our Spring application, we used

JUnit v. 5 to perform unit tests over the developed

code base. In addition, deployment of each of the

client applications was achieved thanks to Tomcat ap-

plication server, like for QADPT.

Finally, given the need to store information about

the user (such as their credentials) and other infor-

mation related to the implemented functionalities, a

MongoDB database was installed and used for the

purpose.

4.2.1 A Cloud-Based, Client-Server Solution

In order to reach the data at each of the federated

nodes, QADAT employs Java RMI technology to

transport result data (i.e., aggregations) from one ma-

chine to another. In addition, and since data could

potentially be large, conveying them over the net-

work requires the usage of specialized libraries such

as RMIIO to enable large data to be transported over

the network in a convenient way. More speciﬁcally,

a Spark running instance server receives the remote

method invocation request from the client through

an RMI call, processes the dataset accordingly (in

a MapReduce fashion) over the Hadoop cluster, and

then returns the results to the client. The results are

then gathered, structured and displayed to users.

4.3 Main Components and

Functionalities

QFLS is intended to satisfy basic analytical needs by

performing analytical computations over diverse data

sources in a federated context. First, QFLS enables

data exploration throughout its Data Federation dis-

covery (DFD). DFD exposes, based on the conﬁgura-

tion ﬁle, the content of each of the federated node in

terms of datasets. Second, the Anonymized Dataset

Analysis (ADA) component provides an analysis of

each of the target anonymized dataset, by providing

details on its level of anonymization of the attributes,

and other statistics such as the minimum value, the

maximum value, the type. Third, the Anonymized

Analytics Environment (AAE) allows users to sub-

mit TLAQ queries onto the remote target datasets and

to retrieve analytics in a tree-like shaped analytical

structure through. QFLS is also capable of deriv-

ing a dashboard over the query answers in order to

enable recommendations based on the obtained ana-

lytics through the Predictive Analytical Environment

(PAE). Furthermore, QFLS enables SQL queries to

be submitted and executed on a target dataset, and

to retrieve the corresponding result set in an intu-

itive interface, easily navigable by non-medical per-

sons, through the SQL Query Environment (SQE).

Each submitted tree-like analytical query is stored in

a database. All of the aforementioned modules use

Java RMI to request and retrieve results from the re-

mote Spark-instance running server.

Figure 3 shows the UML package diagram of

QFLS, as to highlight the inter-dependencies among

the various components.

Figure 3: UML Package Diagram of QFLS.

Since, within the scope of the QUALITOP project,

medical data at the federated nodes cannot be ac-

cessed, all of the data processing logic occurs on

server side (on each federated node) and only ag-

gregate information is returned back to the QADAT

component. This ensures the necessary level of

anonymization, as per the requirements of the QUAL-

ITOP project.

To better understand the data access constraint, we

adopt the strategy of retrieving distinct values within

DATA 2023 - 12th International Conference on Data Science, Technology and Applications

426

dataset columns, which are necessary at various sub-

components of QADAT (e.g., AAE component). In

this case, Spark processing yields only distinct val-

ues through adequately processing the anonymized

dataset on the corresponding node, while data are

never returned raw, nor are fully-processed at the

client side. Indeed, even though data are anonymized

on each federated node, data access may still expose

certain sensitive information such as patient names or

their diseases (e.g., (Sweeney, 2002)). In addition to

this measure, identiﬁer columns are removed at the

time of processing the dataset at the federated node,

as to reinforce anonymity of data and to provide op-

timal security guarantees over the privacy of patients

and data owners.

Finally, QFLS is obviously extendable to a fed-

erated machine learning system (e.g., (Yang et al.,

2019; Fu et al., 2022)), where the processing at each

node level would be based on a machine learning

model rather than a data aggregation model, thus re-

sulting in a global machine model. Under this as-

sumption, SparkML would be our best bet to employ

for embedding further capabilities to our framework

(e.g., (Omran et al., 2021; Mohamed et al., 2021)).

5 CONCLUSIONS AND FUTURE

WORK

This paper has described in details deﬁnitions, re-

quirements, models and techniques of QFLS, a

Cloud-based framework for supporting big health-

care data management and analytics from big data

lakes. QFLS anatomy and main functionalities have

been described, along with the main software solu-

tions proposed with the framework.

Concepts and guidelines deriving from the pro-

posed framework also opens the door to emerging

research challenges for the future. Among those,

an interesting research line consists in extending our

framework as to deal with advanced machine learn-

ing techniques, such as those experimented in other

research efforts (e.g., (Cuzzocrea et al., 2017; Leung

et al., 2019; Coronato and Cuzzocrea, 2022; Liu et al.,

2023a; Adeoye et al., 2023; Tan et al., 2023)).

ACKNOWLEDGEMENTS

This research has been funded by the EU H2020

QUALITOP research project - Call Reference: H2020

- SC1-DTH-01-2019; Project Number: 875171.

REFERENCES

Abbasi, A. and Mohammadi, B. (2022). A clustering-

based anonymization approach for privacy-preserving

in the healthcare cloud. Concurr. Comput. Pract. Exp.,

34(1).

Adeoye, J., Koohi-Moghadam, M., Choi, S., Zheng, L., Lo,

A. W. I., Tsang, R. K., Chow, V. L. Y., Akinshipo,

A., Thomson, P., and Su, Y. (2023). Predicting oral

cancer risk in patients with oral leukoplakia and oral

lichenoid mucositis using machine learning. J. Big

Data, 10(1):39.

Balbin, P. P. F., Barker, J. C. R., Leung, C. K., Tran,

M., Wall, R. P., and Cuzzocrea, A. (2020). Predic-

tive analytics on open big data for supporting smart

transportation services. Procedia Computer Science,

176:3009–3018.

Beaulieu, E., Spanjaart, A., Roes, A., Rachet, B., Dalle, S.,

Kersten, M. J., Maucort-Boulch, D., and Jalali, M. S.

(2022). Health-related quality of life in cancer im-

munotherapy: a systematic perspective, using causal

loop diagrams. Qual Life Res., 31(8).

Campan, A., Cuzzocrea, A., and Truta, T. M. (2017). Fight-

ing fake news spread in online social networks: Actual

trends and future research directions. In IEEE Big-

Data, 2017, pages 4453–4457. IEEE Computer Soci-

ety.

Cannataro, M., Cuzzocrea, A., and Pugliese, A. (2001). A

probabilistic approach to model adaptive hypermedia

systems. In Proceedings of the First International

Workshop on Web Dynamics, WebDyn@ICDT 2001,

London, UK, January 3, 2001, pages 50–60.

Cannataro, M., Cuzzocrea, A., and Pugliese, A. (2002).

XAHM: an adaptive hypermedia model based on

XML. In Proceedings of the 14th international con-

ference on Software engineering and knowledge engi-

neering, SEKE 2002, Ischia, Italy, July 15-19, 2002,

pages 627–634.

Chaudhuri, S. (2012). What next?: a half-dozen data man-

agement research goals for big data and the cloud. In

Benedikt, M., Kr

otzsch, M., and Lenzerini, M., edi-

tors, ACM SIGMOD-SIGACT-SIGART PODS, 2012,

pages 1–4. ACM.

Chen, D., Chen, H., Jiang, Z., and Zhao, Y. (2017). An

adaptive memory tuning strategy with high perfor-

mance for spark. Int. J. Big Data Intell., 4(4):276–

286.

Chen, J., Chen, Y., Du, X., Li, C., Lu, J., Zhao, S., and

Zhou, X. (2013). Big data challenge: a data manage-

ment perspective. Frontiers Comput. Sci., 7(2):157–

164.

Coronato, A. and Cuzzocrea, A. (2022). An innovative risk

assessment methodology for medical information sys-

tems. IEEE Trans. Knowl. Data Eng., 34(7):3095–

3110.

Cuzzocrea, A. (2021). Big data lakes: Models, frameworks,

and techniques. In IEEE BigComp, 2021, pages 1–4.

IEEE.

Cuzzocrea, A., Furfaro, F., Mazzeo, G. M., and Sacc

a, D.

(2004). A grid framework for approximate aggregate

QFLS: A Cloud-Based Framework for Supporting Big Healthcare Data Management and Analytics from Big Data Lakes: Deﬁnitions,

Requirements, Models and Techniques

427

query answering on summarized sensor network read-

ings. In OTM 2004 Workshops, 2004, volume 3292,

pages 144–153.

Cuzzocrea, A., Leung, C. K., and MacKinnon, R. K. (2014).

Mining constrained frequent itemsets from distributed

uncertain data. Future Gener. Comput. Syst., 37:117–

126.

Cuzzocrea, A., Martinelli, F., Mercaldo, F., and Vercelli,

G. V. (2017). Tor trafﬁc analysis and detection via

machine learning techniques. In IEEE BigData, 2017,

pages 4474–4480. IEEE Computer Society.

Dash, S., Shakyawar, S. K., Sharma, M., and Kaushik, S.

(2019). Big data in healthcare: management, analysis

and future prospects. J. Big Data, 6:54.

Dimitsaki, S., Gavriilidis, G. I., Dimitriadis, V. K., and Nat-

siavas, P. (2023). Benchmarking of machine learning

classiﬁers on plasma proteomic for COVID-19 sever-

ity prediction through interpretable artiﬁcial intelli-

gence. Artif. Intell. Medicine, 137:102490.

Ding, J., Errapotu, S. M., Guo, Y., Zhang, H., Yuan, D.,

and Pan, M. (2022). Private empirical risk minimiza-

tion with analytic gaussian mechanism for healthcare

system. IEEE Trans. Big Data, 8(4):1107–1117.

Elgammal, A. and Kr

amer, B. J. (2021). A reference ar-

chitecture for smart digital platform for personalized

prevention and patient management. In Next-Gen

Digital Services. A Retrospective and Roadmap for

Service Computing of the Future - Essays Dedicated

to Michael Papazoglou on the Occasion of His 65th

Birthday and His Retirement, volume 12521, pages

88–99.

Fu, X., Zhang, B., Dong, Y., Chen, C., and Li, J. (2022).

Federated graph machine learning: A survey of con-

cepts, techniques, and applications. SIGKDD Explor.,

24(2):32–47.

Ghayvat, H., Pandya, S., Bhattacharya, P., Zuhair, M.,

Rashid, M., Hakak, S., and Dev, K. (2022). CP-

BDHCA: blockchain-based conﬁdentiality-privacy

preserving big data scheme for healthcare clouds and

applications. IEEE J. Biomed. Health Informatics,

26(5):1937–1948.

Gounaris, A. and Torres, J. (2018). A methodology for

spark parameter tuning. Big Data Res., 11:22–32.

Leung, C. K., Cuzzocrea, A., Mai, J. J., Deng, D., and

Jiang, F. (2019). Personalized deepinf: Enhanced so-

cial inﬂuence prediction with deep learning and trans-

fer learning. In IEEE BigData, 2019, pages 2871–

2880. IEEE.

Lin, C., Song, Z., Song, H., Zhou, Y., Wang, Y., and

Wu, G. (2016). Differential privacy preserving in big

data analytics for connected health. J. Medical Syst.,

40(4):97:1–97:9.

Liu, C., Yao, Z., Liu, P., Tu, Y., Chen, H., Cheng, H., Xie,

L., and Xiao, K. (2023a). Early prediction of MODS

interventions in the intensive care unit using machine

learning. J. Big Data, 10(1):55.

Liu, X., Hasan, M. R., Ahmed, K. A., and Hossain, M. Z.

(2023b). Machine learning to analyse omic-data for

COVID-19 diagnosis and prognosis. BMC Bioinform.,

24(1):7.

Mohamed, M. A., El-Henawy, I. M., and Salah, A.

(2021). Usages of spark framework with different ma-

chine learning algorithms. Comput. Intell. Neurosci.,

2021:1896953:1–1896953:7.

Omran, N. F., Ghany, S. F. A., Saleh, H., and Nabil, A.

(2021). Breast cancer identiﬁcation from patients’

tweet streaming using machine learning solution on

spark. Complex., 2021:6653508:1–6653508:12.

Onesimu, J. A., Karthikeyan, J., Eunice, J., Pomplun, M.,

and Dang, H. (2022). Privacy preserving attribute-

focused anonymization scheme for healthcare data

publishing. IEEE Access, 10:86979–86997.

Parimanam, K., Lakshmanan, L., and Palaniswamy, T.

(2022). Hybrid optimization based learning technique

for multi-disease analytics from healthcare big data

using optimal pre-processing, clustering and classiﬁer.

Concurr. Comput. Pract. Exp., 34(17).

Patil, H. K. and Seshadri, R. (2014). Big data security and

privacy issues in healthcare. In IEEE Congress on Big

Data, 2014, pages 762–765. IEEE Computer Society.

QUALITOP (2023). The QUALITOP project.

https://h2020qualitop.liris.cnrs.fr/wordpress/index.

php/project/.

Singh, S., Rathore, S., Alfarraj, O., Tolba, A., and Yoon,

B. (2022). A framework for privacy-preservation

of iot healthcare data using federated learning and

blockchain technology. Future Gener. Comput. Syst.,

129:380–388.

Sun, J. and Reddy, C. K. (2013). Big data analytics for

healthcare. In ACM SIGKDD KDD, 2013, page 1525.

ACM.

Sweeney, L. (2002). k-anonymity: A model for protecting

privacy. Int. J. Uncertain. Fuzziness Knowl. Based

Syst., 10(5):557–570.

Tan, Q., Xu, X., and Liang, H. (2023). Physiological big

data mining through machine learning and wireless

sensor networks. Int. J. Distributed Syst. Technol.,

14(2):1–12.

Vinke, P. C., Combalia, M., de Bock, G. H., Leyrat, C.,

Spanjaart, A. M., Dalle, S., da Silva, M. G., Essongue,

A. F., Rabier, A., Pannard, M., Jalali, M. S., Elgam-

mal, A., Papazoglou, M., Hacid, M.-S., Rioufol, C.,

Kersten, M.-J., van Oijen, M. G., Suazo-Zepeda, E.,

Malhotra, A., Coquery, E., Anota, A., Preau, M., Fau-

vernier, M., Coz, E., Puig, S., and Maucort-Boulch,

D. (2023). Monitoring multidimensional aspects of

quality of life after cancer immunotherapy: protocol

for the international multicentre, observational quali-

top cohort study. BMJ Open, 13(4).

Yang, Q., Liu, Y., Chen, T., and Tong, Y. (2019). Federated

machine learning: Concept and applications. ACM

Trans. Intell. Syst. Technol., 10(2):12:1–12:19.

Zhang, H., Chen, G., Ooi, B. C., Tan, K., and Zhang, M.

(2015). In-memory big data management and pro-

cessing: A survey. IEEE Trans. Knowl. Data Eng.,

27(7):1920–1948.

DATA 2023 - 12th International Conference on Data Science, Technology and Applications

428