Fynex: Work in Progress on a Web-based Approach That Implements a

Hybrid Recommendation System for Preventing and Treating Diseases

based on Eating Disorders

ıter Andr

es Gonz

alez-Daza

and Miguel Alfonso Feij

oo-Garc

ıa

Program of Systems Engineering, Universidad El Bosque, Bogot

a, Colombia

Keywords:

Usability, Healthcare Application, Recommendation System, Machine Learning.

Abstract:

Diseases based on eating disorders have been considered a health issue worldwide. It is worrying due to the

increased complexity of treatments for diseases such as Diabetes, Hypertension, Obesity, and Anorexia. We

present Fynex as a web-based expert system that implements a hybrid recommendation system that supports

Healthcare Professionals and Patients with recommendations on nutritional plans and physical activity. It

provides functionalities to have detailed follow-up and control regarding the evolution of a particular treat-

ment regarding these diseases. This solution poses a decision-making system for Healthcare Professionals and

Patients, fostering the medical processes (i.e., treatments, progress, and decisions) in an integrated and sys-

temic way. We evaluated this tool following a pilot study based on pre-post experimentation, and reported the

ﬁndings and results considering participants’ performance and perceptions, basing our analysis on (0-5) Lik-

ert scales, open-ended and YN responses, regarding their experience interacting with Fynex— analyzing the

users’ perceptions on satisfaction and usability on the web-based application. Our preliminary ﬁndings sug-

gest that Fynex is effectively a friendly user-centered approach that successfully increases medical processes,

healthcare control, and treatment satisfaction.

1 INTRODUCTION

The worldwide population (i.e., the youth population)

usually does not have correct eating habits, which in-

creases the chances of being diagnosed with a disease

based on eating disorders (Candela, 2016). These dis-

eases based on eating disorders such as Diabetes, Hy-

pertension, Obesity, and Anorexia are considered a

global problem, as this could result in possible cardiac

complications, glucose intolerance, and insulin resis-

tance, among others (Thomas et al., 2019). Moreover,

medical controls are not generally customized, mon-

itoring of these diseases is limited, and the lack of

communication with healthcare professionals (HcP)

could become critical beyond the South American

context.

There are numerous solutions with scaffolds for

health applications in this context. For example, My-

Diabetes.diet (Garcia et al., 2001) or DialBetics (Waki

et al., 2014), have a similar approach to treating this

type of diseases. Furthermore, the research work “In-

https://orcid.org/0000-0002-9587-205X

https://orcid.org/0000-0001-5648-9966

ternet of Things based on electronic and mobile health

systems for blood glucose continuous monitoring and

management” (Barata et al., 2019), applies Internet of

Things (IoT) on glucose monitoring to be measured in

real-time to provide patients control over their signs.

However, these solutions restrict the monitoring of the

treatment through single applications, focusing either

on the patient or HcP.

Hybrid Recommender Systems combine different

techniques (i.e., recommendation systems) to produce

outputs to complete or complement their best features

and make better conjoint recommendations (Seth and

Sharaff, 2022). Due to the hybrid technique’s use-

fulness, many recommendation systems have techni-

cally integrated it (e.g., implementation of solutions

in e-commerce). However, unfortunately, there is

a lack of hybrid recommendation systems focused

on medicine, with only 5% adopting this approach

(Danilova and Ponomarev, 2017). Recommendation

systems focused on medicine use different methods,

such as K-Means and association rules. For example,

supervised models deﬁne which drugs have helped

certain patients with speciﬁc symptoms— the super-

vised model is one of the most common methods

González-Daza, B. and Feijóo-García, M.

Fynex: Work in Progress on a Web-based Approach That Implements a Hybrid Recommendation System for Preventing and Treating Diseases based on Eating Disorders.

DOI: 10.5220/0011543300003323

In Proceedings of the 6th International Conference on Computer-Human Interaction Research and Applications (CHIRA 2022), pages 193-200

ISBN: 978-989-758-609-5; ISSN: 2184-3244

193

applied to recommendation systems. However, they

usually point to multiple diseases without delving into

them (Stark et al., 2019).

We present Fynex as a friendly and user-centered

web-based application, that supports HcP and patients

with treatments regarding diseases based on eating

disorders. It has a hybrid recommendation system,

a messaging chat, and a visual system to visualize the

evolution of a patient, among others. Furthermore,

it integrates Artiﬁcial Intelligence (AI), pretending to

support medical on-time decision-making processes.

Also, it focuses on customized treatment follow-up,

which both HcP and patients could use.

This healthcare-based solution seeks to increase

the satisfaction of both HcP and patients regarding

medical treatments and decision-making processes.

Hence, this paper presents Fynex and the result of our

ﬁrst pilot study using the web-based tool. We evalu-

ated the participants’ perception of satisfaction before

and after using the application through questionnaires

to analyze the behavior and impact of the solution pre-

liminary— we followed a pre-post experimentation

approach (Miller et al., 2020). We also describe the

pros and cons of this pilot. We report the preliminary

ﬁndings and results of this pilot study, pretending to

answer and analyze how the satisfaction of Patients

and HcP is impacted regarding the treatment of dis-

eases based on eating disorders, when supported by

an intelligent system based on a hybrid recommenda-

tion system.

Our work contributes to Healthcare Applications

and User-Centered Interaction literature. We evalu-

ate Fynex as an information and communication tech-

nology (ICT) for healthcare and medicine, customiz-

ing HcP and Patient’s experience to support decision-

making processes (i.e., home care monitoring, or di-

agnostic support) regarding treatments, and on-time

medical control. This approach uses the beneﬁts of

autonomous, e-health-based, and customized medical

processes addressing diseases based on eating disor-

ders.

2 FYNEX: DESCRIPTION

We introduce a web-based application called Fynex

that provides a friendly, user-centered, and interactive

Graphical User Interface (GUI) to support HcP and

patients in medical processes regarding eating disor-

ders. Fynex is presented to end-users who are particu-

larly involved in the treatment of one of the following

preliminary diseases: (1) Diabetes, (2) Hypertension,

(3) Obesity, and (4) Anorexia Nervosa. Therefore, the

platform provides features and functionalities to im-

prove the satisfaction perceived by these actors (i.e.,

HcP and patients), by providing a visually attractive

GUI that eases the user’s interaction on the applica-

tion, getting the most out of it. It implements a hy-

brid recommender system (C¸ ano and Morisio, 2017)

to present possible nutrition and exercise plans for the

patient to follow and apply them in their life (See Fig-

ure 1). Besides, the web-based tool presents a mon-

itoring system to keep track of the patient’s evolu-

tion throughout time. This resulting information is

helpful to the patient as to its HcP. Shneiderman’s

mantra (Munzner, 2014) presents how a tool is in-

tended to manage the information visually through:

(1) overview ﬁrst, (2) zoom and ﬁlter, and (3) de-

tails on-demand, as required by a user (i.e., HcP and

patients). Hence, Fynex provides a dashboard with

health-based variables related to the patient’s treat-

ment, a system to upload and download test results,

a messaging system between the HcP and its patients,

and a visual representation of the similarities between

the patients, among others (See Figure 2).

Fynex is a monolithic application using Django

as its web framework, integrating a Model-View-

Template architecture (MVT) (Rull et al., 2009), and

Angular JS (Green and Seshadri, 2013) as a fron-

tend framework to improve the dynamism and qual-

ity of the web application. Additionally, the ap-

plication integrates third-party tools such as Scikit-

learn (Kramer, 2016), Redis (Carlson, 2013), and

IBM Cloud’s Object Storage service (Samp

e et al.,

2018), to support the preprocessing of the information

and the development of the web-based tool’s manage-

ment, as of its hybrid recommender system (C¸ ano and

Morisio, 2017). The data processing for the infor-

mation system (i.e., Fynex) supports the recommen-

dation models’ implementation to enhance its results

and reinforce further decision-making processes of a

user. Fynex implements an ETL (Vassiliadis and Sim-

itsis, 2009) to process the information from the Fynex

database and be able to use the already trained Ma-

chine Learning (ML) models. Once the data is trans-

formed, and loaded, we used a hybridization method

to obtain a ﬁnal recommendation, which we present

to the user through the application— memory-based

recommendation model, content-based recommenda-

tion model, and the ML model-based recommenda-

tion model (see Figure 3).

3 DATA ACQUISITION

The participants who gathered the experimentation

voluntarily had roles of HcP and Patients. In this pilot

study, we had a total of 14 participants (N=14)— 50%

CHIRA 2022 - 6th International Conference on Computer-Human Interaction Research and Applications

194

Figure 1: Features/Functionalities of Fynex - Part 1.

(n=7) HcP, and 50% (n=7) patients. We had 64.29%

(n=9) male participants (i.e., 85.7% (n=6) HcP, and

42.9% (n=3) Patients), and 35.1% (n=5) female par-

ticipants (i.e., 14.3% (n=1) Hcp, and 57.1% (n=4) Pa-

tients).

Each participant performed speciﬁc tasks in the

application. At the end of the tasks, the participants in

the experimentation phase had to ﬁll out a question-

naire that contained questions regarding the applica-

tion and how much they recommended Fynex.

We evaluated the tool’s effectiveness, usability,

and acceptance, considering the participant’s percep-

tions when using Fynex. We followed pre-post ex-

perimentation (Miller et al., 2020) based on two

questionnaires [quantitative— 0-5 Likert scales (Joshi

et al., 2015), and qualitative— Y/N and Open-Ended

Questions], which were responded before and after

interacting with the web-based tool (See Table 1).

To the evaluation of usability, we applied an ad-

ditional questionnaire based on SUS— System Us-

ability Scale [quantitative— 1-5 scales] (Grier et al.,

2013). Moreover, we asked the participants to answer

the NPS— Net Promoter Score [quantitative— Per-

centage from 0-100%] (Mandal, 2014) (See Table 2).

4 EXPERIMENTAL APPROACH

We sought to understand the behavior change of the

satisfaction perceived by HcP and Patients, before and

after the implementation of Fynex. To understand the

context involved, we conducted our research on eat-

ing disorders and the diseases they might cause (i.e.,

Diabetes, Hypertension, Obesity, and Anorexia).

We considered in our approach the Biopsy-

chosocial and Cultural Model (BPSCM) (Cruz and

Buitrago, 2017) to understand the context based on

Fynex: Work in Progress on a Web-based Approach That Implements a Hybrid Recommendation System for Preventing and Treating

Diseases based on Eating Disorders

195

Figure 2: Features/Functionalities of Fynex - Part 2.

the identiﬁcation of (1) artifacts, (2) means, (3) habits,

and (4) beliefs involved, from each participant’s per-

spective (i.e., HcP and Patients), helping in the under-

standing and enhancement of complex needs— con-

text based on medical processes. Then, we deﬁned

how these elements might change if the actors used

Fynex, so we can verify if those changes contribute to

fulﬁlling the project’s purpose.

Once the hybrid recommendation system of Fynex

was developed, we deﬁned three phases for testing

and evaluating the model. The ﬁrst phase was fo-

cused on evaluating the recommendation model, us-

ing metrics like the ROC curve (de Ullibarri Galpar-

soro and Fern

andez, 1998), F1-Score (Lipton et al.,

2014), and results of simulations, among others. In

the second phase, each participant (i.e., HcP and Pa-

tients) was asked to answer how the experience and

satisfaction changed (e.g., treatment, support) when

using the tool. We calculated the percentage of the in-

crement or decrement of perceived satisfaction when

interacting with the tool based on the ACSI (Fornell

et al., 1996) (i.e., American Customer Satisfaction In-

dex), as well as the Probit and Logit (Chen and Tsu-

rumi, 2010)— regression model based on indepen-

dent variables such as age and time with treatment.

Moreover, in the third phase, we used Loop11 (Busta-

mante, 2010) to establish the tasks to perform and to

obtain certain metrics (i.e., percentage of completed

tasks, usability calculated by the SUS (Grier et al.,

2013)— System Usability Scale, and the acceptance

calculated by the NPS (Mandal, 2014)— Net Pro-

moter Score).

CHIRA 2022 - 6th International Conference on Computer-Human Interaction Research and Applications

196

Figure 3: Concept design of Fynex.

5 FINDINGS AND RESULTS

We present our ﬁndings and results based on the par-

ticipants’ experience on Fynex reported, following

the experimental methodology— the methodological

process and instruments used, were described in sec-

tions 3 and 4.

In the ﬁrst place, we evaluated the predictive AI

models of Fynex, calculating their precision, accu-

racy, sensitivity, and the F1-Score indicators. In addi-

tion, we complemented each model’s confusion ma-

trix with the calculation of the ROC curve, which al-

lowed knowing the AUC (i.e., Area Under the Curve)

(Fawcett, 2006) to know the probability of a model

detecting the disease in a patient in the context— dis-

eases regarding eating disorders.

The diabetes predictive model offered an 85.7%

probability of detecting the disease in patients. The

predictive model of Arterial Hypertension offered a

probability of 83.2% of detecting the disease in pa-

tients.

Furthermore, over 25 different recommendations

provided by the hybrid recommender system of Fynex

for this pilot study, we confronted the results against

the ideal nutritional plans for each disease. We con-

sidered ideal nutritional plans considering the litera-

ture and the particular context— diseases regarding

eating disorders. Patients treating Diabetes typically

consume 15% to 20% protein, 45% to 50% carbo-

hydrate, and 20% to 35% fat (Gray and Threlkeld,

2015). To treat Arterial Hypertension, there is a

standard diet known as DASH (Dietary Approaches

to Stop Hypertension), based on consuming between

15% and 20% protein, between 52% and 58% car-

bohydrates, and between 25% and 30% fat (Camp-

bell, 2017). On the other hand, the diet to treat

Obesity is based on low-calorie, low-fat, and low-

carbohydrate (Fock and Khoo, 2013). Finally, people

treating Anorexia Nervosa usually consume between

15% and 20% protein, between 60% and 65% carbo-

hydrates, and between 20% and 25% fat (Baskaran

et al., 2017).

Obtaining an F1-Score between 70% and 85%

(i.e., we obtained 77%), we can claim that there is

no case of Overﬁtting or Underﬁtting, as there is no

overtraining of the data when having an evaluation far

from perfect. However, there is a sufﬁciently success-

ful evaluation to deduce that there is also no poor data

training (Ying, 2019). In addition, from the analysis

of the ROC curves, the probability of correctly detect-

ing Diabetes and Hypertension in patients was 85.7%

and 83.2%, respectively, which lets us infer it is a high

Fynex: Work in Progress on a Web-based Approach That Implements a Hybrid Recommendation System for Preventing and Treating

Diseases based on Eating Disorders

197

Table 1: Questionnaire regarding HcP’s and Patient’s pre/post-experience using Fynex.

Question Type of Question:

(1)

, OE

(2)

, LS

(3)

or YN

(4)

Role:

HcP

(5)

or P

(6)

Moment of

Application:

(7)

or A

(8)

Q1: Provide your gender. MC HcP &

B & A

Q2: Provide your age. OE HcP &

B & A

Q3: For how many years have you been a HcP? OE HcP B & A

Q4: How secure do you feel about the medical

recommendations you give to your patients?

LS HcP B & A

Q5: Justify your provided answer in Q4. LS HcP B & A

Q6: How easy do you consider is it to monitor your

patient’s health condition?

LS HcP B & A

Q7: How easy is it to communicate with your

patient/HcP?

LS HcP &

B & A

Q8: How useful are the nutritional and exercise plans

recommended by Fynex, for the treatment of diseases

based on eating disorders?

LS HcP A

Q9: Is the website comfortable and attractive, allowing

easy use of the application?

YN HcP A

Q10: Is the information offered by the application

clear and sufﬁcient?

YN HcP A

Q11: The nutrition and exercise plans

recommendations are generated fast?

YN HcP A

Q12: For how many years have you been a medical

treatment?

OE P B & A

Q13: Do you think you eat well? YN P B & A

Q14: Are you in treatment with a nutrition

professional?

YN P B & A

Q15: How satisﬁed are you with your treatment? LS P B & A

Q16: How much do you trust your HcP’s

recommendations?

LS P B & A

Q17: Justify the provided answer in Q17. OE P B & A

Q18: How easy is it to monitor your own health? LS P B & A

Q19: Is Fynex comfortable and attractive, allowing

easy use of the application?

YN P A

Q20: Is the information provided by Fynex clear and

sufﬁcient?

YN P A

Q21: How easy is it to use Fynex? LS HcP &

Q22: Would you recommend Fynex to others? YN HcP &

Q23: Provide additional comments OE HcP &

B & A

Notes:

(1)

MC: Multiple Choice,

(2)

OE: Open-Ended Question,

(3)

LS: [0-5] Likert Scale,

(4)

: YN:Y/N Questions,

(5)

HcP: Healtcare Professional,

(6)

P: Patient

(7)

B: Before using Fynex,

(8)

A: After using Fynex

probability of preventing and treating these diseases.

We calculated the usability and acceptance tests

through the SUS (i.e., System Usability Scale) and

the NPS ( i.e., Net Promoter Score). As a result, HcP

obtained an average SUS of 76.69, and the patients

obtained a SUS of 90.36. Additionally, the NPS ob-

tained by HcP was 14.29%. Even if the resulting

value was not that signiﬁcant, we considered it a posi-

tive possibly recommendation (Mendieta Giron et al.,

2017). Likewise, patients resulted with an NPS of

CHIRA 2022 - 6th International Conference on Computer-Human Interaction Research and Applications

198

Table 2: Questionnaire to calculate System Usability Scale (SUS) and Net Promoter Score (NPS) of Fynex.

Question Scale Option

I think that I would like to use this system frequently. [1-5] SUS

I found the system unnecessarily complex. [1-5] SUS

I thought the system was easy to use. [1-5] SUS

I think that I would need the support of a technical person to be able to use this system. [1-5] SUS

I found the features and functionalities in this system were well integrated. [1-5] SUS

I thought there was too much inconsistency in this system. [1-5] SUS

I would imagine that most people would learn to use this system very quickly. [1-5] SUS

I found the system very cumbersome to use. [1-5] SUS

I felt very conﬁdent using the system. [1-5] SUS

I needed to learn a lot of things before I could get going with this system. [1-5] SUS

How likely are you to recommend Fynex? [1-10] NPS

85.71%, which is a signiﬁcant value due to the re-

lationship between promoters and detractors.

6 DISCUSSION

From the preliminary results and comments received

by the participants of this pilot study, Fynex effec-

tively provides friendly and user-centered support

to medical processes and healthcare control, regard-

ing diseases based on eating disorders. It integrates

AI models and content-based, ML model-based, and

memory-based recommendation systems supporting

nutrition and physical activity (i.e., medical treatment

control or process). Regardless of the results’ volatil-

ity, satisfaction increased in both roles (i.e., HcP and

Patients). From the preliminary results, both feel

more satisﬁed with the treatment process when using

Fynex as a support tool. Although both roles consider

that the application does offer support, it was more

signiﬁcant in patients than HcP. Hence, we could

claim that patients are more open to new technologies

and trying different solutions to address their prob-

lems more directly. Moreover, the Harris Poll, in as-

sociation with Stanford University, analyzes these cir-

cumstances, particularly HcP satisfaction with health-

focused technologies (Wuerdeman et al., 2005). They

claim that 55% of HcP are willing to experiment and

use new technologies only if other HcP have used

them before and have recommended their use.

The results (i.e., before and after interacting with

Fynex) show that the HcP’s satisfaction increased by

32.95% in the experiment. However, when applying

the Probit and Logit technique, the satisfaction in-

creased by 57.6%. On the other hand, the patients’

satisfaction increased by 56.96% in the experiment.

Also, the increase obtained when comparing the pre-

dictions of the patients’ models used in the Probit and

Logit was 55.47%.

We ﬁnd that a web-based tool such as Fynex

supports and complements medical processes (i.e.,

healthcare control, diagnoses, and treatments) for

HcP and Patients. Based on the participants’ com-

ments regarding Fynex, in addition to the different

similar tools that exist, the participants found sat-

isfactory the possibility that both HcP and patients

could interact in the medical treatment processes syn-

chronously on a single platform. Fynex allows both

end-users (i.e., HcP and patients) to have on-time in-

formation regarding the available treatment process

independently. HcP can monitor their patients and

make decisions to support medical treatments. Like-

wise, patients can access and monitor their health sta-

tus (i.e., healthcare control), hand in hand with the

HcP medical control and decisions. In further re-

search, a deepened analysis will be addressed to com-

plement the preliminary claims presented in this pa-

per.

ACKNOWLEDGEMENTS

The authors thank Dayan Alejandra Hincapi

e-Cort

for her dedication, performance, and dedication to the

successful development of Fynex, which was vital to

achieving the results and ﬁndings obtained and re-

ported in this pilot study. Also, the authors thank all

participants who voluntarily and actively contributed

to the preliminary experimentation of Fynex.

REFERENCES

Barata, J. J. R., Munoz, R., Silva, R. D. D. C., Rodrigues,

J. J., and De Albuquerque, V. H. C. (2019). Internet of

things based on electronic and mobile health systems

for blood glucose continuous monitoring and manage-

ment. IEEE Access, 7:175116–175125.

Fynex: Work in Progress on a Web-based Approach That Implements a Hybrid Recommendation System for Preventing and Treating

Diseases based on Eating Disorders

199

Baskaran, C., Carson, T. L., Campoverde Reyes, K. J.,

Becker, K. R., Slattery, M. J., Tulsiani, S., Eddy, K. T.,

Anderson, E. J., Hubbard, J. L., Misra, M., et al.

(2017). Macronutrient intake associated with weight

gain in adolescent girls with anorexia nervosa. In-

ternational Journal of Eating Disorders, 50(9):1050–

1057.

Bustamante, J. (2010). La herramienta de tests de usabili-

dad a distancia loop11. Profesional de la Informacion,

19(4):425–430.

Campbell, A. P. (2017). Dash eating plan: an eating pat-

tern for diabetes management. Diabetes Spectrum,

30(2):76–81.

Candela, J. M. (2016). ¿ cu

ales son los factores de riesgo

para desarrollar diabetes mellitus tipo 2? GU

IA DE.

C¸ ano, E. and Morisio, M. (2017). Hybrid recommender

systems: A systematic literature review. Intelligent

Data Analysis, 21(6):1487–1524.

Carlson, J. (2013). Redis in action. Simon and Schuster.

Chen, G. and Tsurumi, H. (2010). Probit and logit model

selection. Communications in Statistics—Theory and

Methods, 40(1):159–175.

Cruz, O. L. and Buitrago, C. H. O. (2017). Formaci

on de

ingenieros inform

aticos como gestores de innovaci

on:

Transformadores de vidas, constructores de futuros

posibles. ANFEI Digital, (7).

Danilova, V. and Ponomarev, A. (2017). Hybrid recom-

mender systems: The review of state-of-the-art re-

search and applications. In Proceedings of the 20th

Conference of FRUCT Association.

de Ullibarri Galparsoro, L. and Fern

andez, P. (1998). Cur-

vas roc. Atenci

on Primaria en la Red, 5(4):229–35.

Fawcett, T. (2006). Introduction to receiver operator curves.

Pattern Recognit. Lett, 27:861–874.

Fock, K. M. and Khoo, J. (2013). Diet and exercise in man-

agement of obesity and overweight. Journal of gas-

troenterology and hepatology, 28:59–63.

Fornell, C., Johnson, M. D., Anderson, E. W., Cha, J., and

Bryant, B. E. (1996). The american customer satisfac-

tion index: nature, purpose, and ﬁndings. Journal of

marketing, 60(4):7–18.

Garcia, M. A., Gandhi, A. J., Singh, T., Duarte, L., Shen, R.,

Dantu, M., Ponder, S., and Ramirez, H. (2001). Esdi-

abetes (an expert system in diabetes). In Proceedings

of the twelfth annual CCSC South Central conference

on The journal of computing in small colleges, pages

166–175.

Gray, A. and Threlkeld, R. J. (2015). Nutritional recom-

mendations for individuals with diabetes.

Green, B. and Seshadri, S. (2013). AngularJS. ” O’Reilly

Media, Inc.”.

Grier, R. A., Bangor, A., Kortum, P., and Peres, S. C.

(2013). The system usability scale: Beyond stan-

dard usability testing. In Proceedings of the Human

Factors and Ergonomics Society Annual Meeting, vol-

ume 57, pages 187–191. SAGE Publications Sage CA:

Los Angeles, CA.

Joshi, A., Kale, S., Chandel, S., and Pal, D. K. (2015). Lik-

ert scale: Explored and explained. British journal of

applied science & technology, 7(4):396.

Kramer, O. (2016). Scikit-learn. In Machine learning for

evolution strategies, pages 45–53. Springer.

Lipton, Z. C., Elkan, C., and Narayanaswamy, B. (2014).

Thresholding classiﬁers to maximize f1 score. arXiv

preprint arXiv:1402.1892.

Mandal, P. C. (2014). Net promoter score: a conceptual

analysis. International Journal of Management Con-

cepts and Philosophy, 8(4):209–219.

Mendieta Giron, C. F. et al. (2017). Ventajas del net pro-

moter score para la atenci

on al cliente en las empresas

de telecomunicaciones.

Miller, C. J., Smith, S. N., and Pugatch, M. (2020). Exper-

imental and quasi-experimental designs in implemen-

tation research. Psychiatry Research, 283:112452.

Munzner, T. (2014). Visualization analysis and design.

CRC press.

Rull, G., Farr

e, C., Teniente, E., and Urp

ı, T. (2009). Mvt: a

schema mapping validation tool. In Proceedings of the

12th International Conference on Extending Database

Technology: Advances in Database Technology, pages

1120–1123.

Samp

e, J., Vernik, G., S

anchez-Artigas, M., and Garc

ıa-

opez, P. (2018). Serverless data analytics in the ibm

cloud. In Proceedings of the 19th International Mid-

dleware Conference Industry, pages 1–8.

Seth, R. and Sharaff, A. (2022). A comparative overview

of hybrid recommender systems: Review, challenges,

and prospects. Data Mining and Machine Learning

Applications, pages 57–98.

Stark, B., Knahl, C., Aydin, M., and Elish, K. (2019). A

literature review on medicine recommender systems.

International journal of advanced computer science

and applications, 10(8).

Thomas, R., Halim, S., Gurudas, S., Sivaprasad, S., and

Owens, D. (2019). Idf diabetes atlas: A review

of studies utilising retinal photography on the global

prevalence of diabetes related retinopathy between

2015 and 2018. Diabetes research and clinical prac-

tice, 157:107840.

Vassiliadis, P. and Simitsis, A. (2009). Extraction, transfor-

mation, and loading. Encyclopedia of Database Sys-

tems, 10.

Waki, K., Fujita, H., Uchimura, Y., Omae, K., Aramaki, E.,

Kato, S., Lee, H., Kobayashi, H., Kadowaki, T., and

Ohe, K. (2014). Dialbetics: a novel smartphone-based

self-management support system for type 2 diabetes

patients. Journal of diabetes science and technology,

8(2):209–215.

Wuerdeman, L., Volk, L., Pizziferri, L., Tsurikova, R., Har-

ris, C., Feygin, R., Epstein, M., Meyers, K., Wald,

J. S., Lansky, D., et al. (2005). How accurate is in-

formation that patients contribute to their electronic

health record? In AMIA annual symposium proceed-

ings, volume 2005, page 834. American Medical In-

formatics Association.

Ying, X. (2019). An overview of overﬁtting and its solu-

tions. In Journal of physics: Conference series, vol-

ume 1168, page 022022. IOP Publishing.

CHIRA 2022 - 6th International Conference on Computer-Human Interaction Research and Applications

200