=Paper=
{{Paper
|id=Vol-2028/paper16
|storemode=property
|title=Recommendation System based on CBR algorithm for the Promotion of Healthier Habits
|pdfUrl=https://ceur-ws.org/Vol-2028/paper16.pdf
|volume=Vol-2028
|authors=Gineth M Cerón-Rios,Diego M Lopez-Gutierrez,Belén Díaz-Agudo,Juan A. Recio-García
|dblpUrl=https://dblp.org/rec/conf/iccbr/Ceron-RiosGDR17
}}
==Recommendation System based on CBR algorithm for the Promotion of Healthier Habits==
https://ceur-ws.org/Vol-2028/paper16.pdf
167
Recommendation System based on CBR
algorithm for the Promotion of
Healthier Habits
Gineth M Cerón-Rios, Diego M Lopez-Gutierrez,
Belén Dı́az-Agudo, and Juan A. Recio-Garcı́a
Department of Telematic Engineering
Universidad del Cauca Popayán, Colombia
email:gceron@unicauca.edu.co, dmlopez@unicauca.edu.co
Department of Software Engineering and Artificial Intelligence
Universidad Complutense de Madrid, Spain
email: belend@ucm.es, jareciog@fdi.ucm.es
Abstract. Recommender systems are becoming very popular as they
are able to predict the preferences of a user. This make recommendation
based on the user profile, past ratings or/and additional knowledge such
as user contextual information. Applied to the health area, they can
take advantage of context information to support health promotion and
disease prevention.
We present a recommender system for the promotion of physical activity
called CoCARE. It recommends videos about physical activity based on a
user profile, his/her context. The main challenge of CoCARE is the small
set of videos to be recommended, because the selection of the videos is
done manually by of health experts. Several health recommender systems
have this same problem. Although today there are a large number of
videos available on the Internet related to physical activity. These could
not be included in the data base of CoCARE; because these do not have
enough information to be categorized and profiled.
This article proposes a CBR system, this assigns a physical activity cat-
egory to new video. In this way the new video will be added to the list
of CoCARE recommendations. In this CBR process, basically consists
on analyzing the description of the new video and compare it with the
cases base of CoCARE, selecting the category of most similar cases.
1 Introduction
Recommender systems in the health area have been proven as useful tools to
help patient-oriented decision making systems, promoting physical activity and
disease prevention, in general, to improve health conditions through healthier
habits. Health recommender systems (HRS) aim to promote health programs, to
provide patients with relevant information, products or services, using knowledge
about his/her personal health record systems [1].
In the literature, we find few systems that recommend health educational
multimedia contents [2–6]. Users get recommendation of exercises (stretching,
Copyright © 2017 for this paper by its authors. Copying permitted for private and
academic purpose. In Proceedings of the ICCBR 2017 Workshops. Trondheim, Norway
168
strengthening,etc.), with outdoor or indoor sessions, based on the user informa-
tion taken from mobile devices, activity bracelets, sensors, and his/her personal
health records and risk factors [7].
We have developed “CoCARE” a platform for promotion of healthy lifestyle
on the basis of a context-aware recommendation system designed for mobile
smart devices [8]. Advancements in technology, mobile devices, sensors, and
wearable devices, provide users with self-monitoring dynamically acquired in-
formation of her physical activities. CoCARE recommends multimedia content
of physical activity and healthy diet based on a user-context model. Given a
user profile and a category, the system recommends some videos about conve-
nient physical activities for this user at this moment. Our system relies on an
initial database of activity videos that are labeled with information used dur-
ing the recommendation process. Currently the system has a limited number of
videos that have been manually acquired from experts in the health area.
CoCARE has a database with 80 videos. These have been tagged with its
title, description, category and keywords (see example in Table 1). One video
could be recommended to several users based on a decision model given by
domain expert.
Title Description Category Keywords
Physical Rehab and Revive Physical Therapy We can and we Walk advance,
Therapist will get better together! Orange County Physical Thera- amble,
Shows pist and Certified Functional Manual Therapist, Dr. Lin foot it
How To talks about how the hip, the legs,and the arms correlate advance,
Walk to proper walking and how it can help you walk more amble,
Correctly efficiently. Proper walking helps prevent pain and other foot it.
chronic injuries.
Table 1: Example of CoCARE Videos Description
Concretely, the decision model is built from a dataset of 597 instances (rows),
6 attributes and 1 main class (see Table 2) created by the expert. CoCARE builds
a decision tree using a supervised learning algorithm. Then, the system classifies
the query with information about the user and his/her context and uses the tree
to recommend contents based on the current user situation [9].
In this paper we deal with the problem of video acquisition and tagging.
Internet provides with a huge amount of videos, most of free use, related with
physical activities: dancing, running, fitness, GAP. Our goal is to use these videos
as recommendation items in our system. To do that, we would need to annotate
the videos with information about the potential users that would benefit from
them. We propose a CBR system to automatically classify videos given its textual
description. This CBR system also computes similarity between the CoCARE
user profiles set and the new video categories, to found its categories.
The paper runs as follows. Section 2 describes the recommender system of
CoCARE based on decision model. Section 3 explains the CBR process to auto-
169
matically annotate new videos. Section 4 evaluates the CBR system. Section 6
concludes the paper and discusses some lines of future work.
2 CoCARE
CoCARE (see Figure 1) is a context aware recommender system that recom-
mends videos on physical activity (PA) and healthy diet (HD) to patients for
promotion of her healthy habits. CoCARE incorporates a context- adaptable
interface based on decision trees.
Fig. 1: Mobile CoCARE
CoCARE recommends multimedia content of physical activity and healthy
diet based on user and contextual information. The basic user model includes
details on the user personal profile (see Table 2). The system takes advantage
of additional contextual factors to provide with personalized recommendations
of multimedia content. The query includes static information like user profile,
and dynamic features like geo-lacation or indoor location, date (day or season),
daily schedule of the user and it can detect when the user has company. [8].
Although the CoCARE system works well as a prototype, it relies on an
initial video database of 80 videos. That means different problems:
– Users get repeated contents after a while.
– Lack of novelty contents provokes user desertion.
– The task of including new videos is cumbersome.
– New videos were included without expert supervision and they were misclas-
sified and never recommended to the right users.
170
We propose a CBR solution to solve these problems and assign tags (video
category, keywords and user profile tags) to new videos based on the comparison
to existing ones.
Attribute Description Value
BMI Represents a previous inference Low weight, normal weight,
in the user’s physical condition. overweight, obesity.
It is a nominal fact type.
Life cycle It is a model fact inferred from Teenagers, adult.
their date of birth. Despite be-
ing an integer, is taken as a
nominal value in the relation-
ships table.
Ethnicity Indicates the racial group a per- It is a nominal value. Indige-
son belongs. nous, afro, other.
Trauma It represents a person with dis- Mobility, visual, auditory, with-
ability. It is a nominal fact. out trauma.
Preference It is the aim of the system user. Health, beauty, sport.
It is a nominal fact.
Cardiovascular disease Indicates a user clinical condi- Diabetes, hypertension, without
tion. It is a nominal fact. risk.
Category It is the class of dataset. It is a Dance, walk, bodily exercises,
nominal fact. stretch, stretch eyes, limbs, per-
sonal hygiene, HIIT, labours,
labours limbs, labours eyes,
LISS, swim, eyes,relaxation,
SCC, jog.
Table 2: User Profile Attributes
3 CBR process
Figure 2 shows the CBR process. To classify new videos, we have implemented
two sequential CBR systems. The first CBR system (CBR1) receives the de-
scription of new video and returns categories from similar videos. The case base
is gathered from the CoCARE video data base and contains 80 instances. Each
case is described by 6 keywords and its solution is a category tag (see categories
in table 2). For example:
– Description = prancing, tapping, dribbling, moving, braiding, waltz.
– Solution = dancing.
This CBR1 module implements a k Nearest Neighbour algorithm to find the
most suitable categories for a given video description. Concretely, we use a
3-NN algorithm with a keyword based similarity measure to select the three
categories with highest similarity values.
171
Fig. 2: CBR Process
Once the categories have been retrieved from the first case base, the sec-
ond CBR module(CBR2) estimates the most suitable user profiles for the video
description.
This second module has a case base with 597 instances where every case is
described by several categories and 6 user profile attributes as the solution (see
table 2) . By this way, the solution will be a new user profile UP. The algorithm
compares locally the similarity value for every attribute of the user profile (
see table 2) using majority voting or weighted majority voting to select the fit
attribute to the solution.
CB 1 = < C11 , C21 , . . . , Cm
1
> (1)
Ci1 = < keywords, categories > (2)
CB 2 = < C12 , C22 , . . . , Cn2 > (3)
Ci2 = < categories, U P > (4)
UP = < BM I, age, et, tr, pr, mc, category > (5)
4 Evaluation
We evaluate our system using leave-1-out cross-validation.
We used the video description of each case on CB as a query. We proposed
the category and profile for U Pr , that is compared to the stored solution U Pq .
We compute the similarity between attributes of U Pq and U Pr , using a binary
function [0,1]. We compared if the attributes of the retrieved profile U Pr are
equal to attributes of test case U Pq . So we calculated the α value (see the
equation ec. 6)
172
Sim(U Pq , U Pr ) = α
where
a = Sim(BM Iq , BM Ir )[0, 1]
b = Sim(ageq , ager )[0, 1]
c = Sim(etq , etr )[0, 1]
d = Sim(trq , trr )[0, 1]
e = Sim(prq , prr )[0, 1]
f = Sim(mcq , mcr )[0, 1]
g = Sim(categoryq , categoryr )[0, 1]
α = 0.1 ∗ (a + b + c + d + e + f ) + 0.4 ∗ g (6)
(7)
The table 3 shows an example. Our query in this example is U Pq and D=
“Steve and Jackie take you through how to get the most out of power walking
and show you how beneficial it truly is. Yes it is an Olympic sport!”.
First the CBR1 module got 3NN categories as: Walk, Exercises and HIIT.
So the CBR2 module compared the UP associated to these categories and found
the most similarity U Pr . Next the system uses cross validation and retrieves
a success solution only if the similarity measure α ≥ 0.7. This process is the
comparison between attributes U Pq and U Pr . For example the table 3 shows a
test case, we obtained a score of α greater than 0.7, so U Pr was added to CB.
Test Case BMI age et tr pr cv V
U Pq normal weight Adult other without trauma beauty diabetes Walk
U Pr overweight Young Other without trauma beauty diabetes Walk
Test Case a b c d e f g
Test Case 0 0 1 1 1 1 1
Table 3: Example cross validation
In each leave-1-out step, we obtained 3 values: the similarity of the best cases
returned by the CBR1 module (1-NN), the 2 best cases (2-NN) and the 3 best
cases (3-NN). Next we made 2 tests with Majority Voting (MV) and Weighted
Majority Voting (WMV) in CBR2 module. Figure 3 shows results from our
experiment as:
– Case 1, it is represented by the blue bar. We found the user profile with
majority voting (MV) for 1NN, 2NN and 3NN.
– Case 2, it is represented by the red bar. We found the user profile with
weighted majority voting (WMV) only for 2NN and 3NN.
173
We conclude that the better result from our CBR system was 3NN with
WMV. In this case, we obtained a value of α greater than 0.85 surpassing the
results achieved of the other tests. Our experiment shows that greater than 90%
of the cases are correctly classified.
Fig. 3: Average of similarity α
5 Related works
The literature to continuation are about design and implementation of CoCARE.
We approached works of health recommendation systems.
Related works such as [3, 4, 6, 10, 11] are mobile context platforms that in-
tegrates sensor technology, cognitive tutoring and evidence-based social design
for health promotion. User selects group activities (jogging, walking, fitness, and
yoga) to make recommendations about stretching exercises, outdoor strengthen-
ing or others, based on gender, age, weight, height, location. Diabeticlink [5], is
a mobile recommender of videos and articles about exercise and healthy diabetic
diet, based on user data and data of sensors. It uses the collaborative filtering
recommendation technique. Finally, it generates progress reports based on the
user blood glucose, his/her lifestyle, body mass index and time of physical activ-
ity. Kalico [12] is a mobile recommender system of healthy food restaurants, the
user suggestion are based in his/her profile, location, budget and preferences. It
provides a list of nearby restaurants in alphabetical order and presents a list of
healthy menus recommendation in each it. Kalico is a system that only promotes
healthy eating and for people who want to eating out.
The previously works mention the use of user models, data modeling and
use of recommendation techniques, but they do not describe the selection of
174
recommendation techniques and how performance the validation them. On the
other hand, these works didn‘t make mention about how many recommendations
have their systems, apparently the systems has few contents to recommender.
The next related works relate with CBR topic. [13, 14], these works use a
CBR algorithm to recommend diabetes care videos to adults, however there is
no evidence that their systems can retrieves additional information from the
videos description. There are other systems that retrieve textual information
[10, 15, 16], recover the sentences that are necessary and complete the sentence,
others recover symptoms of some disease with the user profile. However, there
is no evidence a system that finds user profiles with just the description of an
item (video).
Our CBR systems retrieves a user profile from the video description. This
could be useful in others areas such as education, commerce and / or advertising.
For example one recommendation systems could find a user profile fit for learning
content or advertising videos with just the description of item.
6 Discussion and Conclusions
We have described our CoCARE recommender system. CoCARE recommends
videos of physical activity categorized by health experts. But the problem is
that they are very few, to include a new video must be properly categorized
for a user profile. Our CBR allows you to categorize the video and find an
appropriate profile from the description of a video. In this work we proposed a
system composed of 2 CBR system, the first categorizes the new video and the
second delivers the appropriate profile. We have evaluated that the CBR system
delivers a better response if the first CBR is 3NN and CBR2 is with similarity.
Our CBR system uses little input knowledge to get an adequate solution. It
offers a simpler alternative to associate videos to the needs and preferences of
different users.
Our system benefits the user and the health expert, with the possibility of
having new recommendations that help the adherence of the physical activity
program.
The work presented in this paper opens several lines of future work.
When you have very short video descriptions the CBR system loses precision
in finding the right category, although the results we obtained are very promising
we have considered that they can be improved if we extend the description from
synonyms using an ontology of synonyms and algorithms matching of learning.
We plan to take information about the most viewed videos on the Internet
(YouTube) and use their description to classify them, assign to the new video an
appropriate user profile and add to CoCARE case base CB automatically using
collaborative filtering and CBR.
175
7 Acknowledgments
This work was performed under the doctoral thesis “Context-Aware Recom-
mender System to Physical Activity Promotion” financed by Colciencias, un-
der call “Convocatoria No 6172 (Doctorados Nacionales)”. Supported by UCM
(Group 921330) and Spanish Committee of Economy and Competitiveness (TIN2014-
55006-R).
References
1. Wiesner, M., Pfeifer, D.: Health recommender systems: concepts, requirements,
technical basics and challenges. International journal of environmental research
and public health 11(3) (2014) 2580–2607
2. Adomavicius, G., Mobasher, B., Ricci, F., Tuzhilin, A.: Context-Aware Recom-
mender Systems. AI MAGAZINE 32(4) (2011) 217–253
3. F, B., L., C.: Mopet: a context-aware and user-adaptive wearable system for fitness
training.. Artificial Intelligence in Medicine 42(2) (2008) 153–163
4. Lamprinos, I., Papadaki, C.: Mobile personal health application for empowering
diabetic patients: The case within empower project. In: Proceedings of The Inter-
national eHealth Telemedicine and Health ICT Forum. Luxembourgh. (2013)
5. Chen, H., Compton, S., Hsiao, O.: Diabeticlink: A health big data system for
patient empowerment and personalized healthcare. In: Smart Health, Springer
(2013) 71–83
6. Du, H., Youngblood, G.M., Pirolli, P.: Efficacy of a smartphone system to support
groups in behavior change programs. In: WH ’14 Proceedings of the Wireless
Health 2014 on National Institutes of Health, ACM (2014) 1–8
7. Mettouris, C., Papadopoulos, G.A.: Contextual Modelling in Context-Aware Rec-
ommender Systems: A Generic Approach. Web Information Systems Engineering
WISE 2011 (2013) 41–53
8. González, K.X., Carvajal, M.Á., Cerón, G.M., López, D.M.: Modelado de un
sistema consciente del contexto para soportar intervenciones en actividad fı́sica y
nutrición saludable. Revista Ingenierı́as Universidad de Medellı́n 15(28) (2016)
9. Girón, E., Cerón, G.M., Rico, C., López, D.M.: Framework for data model to
personalized health systems. In: 2016 8th Euro American Conference on Telematics
and Information Systems (EATIS). (2016) 1–5
10. El-Sappagh, S.H., El-Masri, S., Elmogy, M., Riad, A.M., Saddik, B.: An ontological
case base engineering methodology for diabetes management. Journal of Medical
Systems 38(8) (2014) 67
11. Cochrane, K., Schiphorst, T.: Developing design considerations for mobile and
wearable technology: M-health applications that can support recovery in mental
health disorders. In: Proceedings of the 9th International Conference on Perva-
sive Computing Technologies for Healthcare. PervasiveHealth ’15, ICST, Brussels,
Belgium, Belgium, ICST (Institute for Computer Sciences, Social-Informatics and
Telecommunications Engineering) (2015) 29–36
12. Anwar, M., Hill, E., Skujins, J., Huynh, K., Doss, C.: Kalico: A smartphone
application for health-smart menu selection within a budget. In: International
Conference on Smart Health, Springer (2013) 113–121
176
13. Bermingham, A., O’Rourke, J., Gurrin, C., Collins, R., Irving, K., Smeaton, A.F.:
Automatically recommending multimedia content for use in group reminiscence
therap. In: Proceedings of the 1st ACM International Workshop on Multimedia
Indexing and Information Retrieval for Healthcare. MIIRH ’13, New York, NY,
USA, ACM (2013) 49–58
14. Polignano, M., de Gemmis, M.: Recommender systems supporting decision making
through analysis of user emotions and personality. In: DC@ AI* IA. (2015) 43–48
15. Linckels, S., Meinel, C.: Applications of description logics to improve multimedia
information retrieval for efficient educational tools. In: Proceedings of the 1st
ACM International Conference on Multimedia Information Retrieval. MIR ’08,
New York, NY, USA, ACM (2008) 321–328
16. Sandhu, R., Kaur, J., Thapar, V.: An effective framework for finding similar cases of
dengue from audio and text data using domain thesaurus and case base reasoning.
Enterprise Information Systems 0(0) (0) 1–18