A Semantic Rules & Reasoning based Approach for ...

A Semantic Rules & Reasoning based Approach for Diet and Exercise Management for Diabetics Irshad Faiz, Hamid Mukhtar, Ali Mustafa Qamar and Sharifullah Khan National University of Sciences and Technology (NUST), Islamabad, Pakistan {10msitiirshad, hamid.mukhtar, mustafa.qamar and sharifullah.khan}@seecs.edu.pk Abstract— Diabetes is a serious chronic disease and balance diet as well as regular exercise are leading important factors for diabetes control. Management of healthy diet and proper exercise involve many decision variables from different domains such as gender, weight, height, age, needed calories and nutrition values, preferences about food and exercise, clinical guidelines and current vital signs etc. We have implemented a semantic rules and reasoning based approach that generates diet and exercise suggestions for diabetics. The solution is developed as a Semantic Healthcare Assistant for Diet and Exercise (SHADE). For each domain (person, diabetes, food and exercise) we have defined separate OWL based ontology along with SWRL rules and then an integrated ontology combines these individual ontologies. Based on data and rules, Pellet reasoner semantically generates the suggestions as inferences. Each generated meal menu is a list of food items along with portion size such that food items are user’s preferred and menu is personalized, healthy and balanced diet. Finally, SHADE recommends user’s preferred activities as exercises along with duration and intensity. Keywords— Diabetes, diet, inference, ontology, OWL, Pellet, reasoner, semantics, SWRL

I. INTRODUCTION

D

iabetes patients increase rate in all over the world and specially in Pakistan is very alarming. Currently more than 88,000 deaths occur in a day due to diabetes and its related complications [1]. Besides medicine, healthy diet, regular exercise, weight control and proper blood glucose test are main controlling factors for diabetes treatment and avoid the complications [2]. Existing solutions for diabetes diet management neither consider integrated knowledge from various domains nor they are using reasoning and inferences based approach. Semantic web [3] is a collection of technologies with the aim that data can be accessed, understood and processed by machines. The idea of semantic web is enabling machines to understand the semantics of data and realizing the concepts and their relationships behind it. Ontology is being used for knowledge representation and Web Ontology Language [4] (OWL) is one of the semantic web supported technology for knowledge expression. Semantic Web Rule Language (SWRL) [5] is the language to facilitate for defining domain knowledge based rules over the concepts and relationships. Reasoning engine is introduced so that machine itself can infer new/additional knowledge by using the rules over the concepts and relationships. Pellet [6] is one of the implementation of the semantic reasoning engine.

Using ontologies and semantics in medical field is very important and advantageous [7]. Ontology is the abstract conceptualization of real world concepts and OWL enables to explicitly define various medical terminologies, concepts and their interrelationships in terms of classes and properties. We have domain rules in the form of clinical guidelines, fitness tips and useful health recommendations. Such rules can be defined to be implemented on the concepts and reasoning can be made to generate new information about the data as inferences. So, an integrated ontology based rules and reasoning is a natural approach towards enabling machine for making intelligent decisions and leading the system towards automation. The objective of this work is to model, design and develop a Semantic Healthcare Assistant for Diet and Exercise (SHADE) that recommends diet and exercise suggestions, initially as a case study, for type 2 diabetes patients. SHADE’s generated recommendations are based on individual’s preferences, personal health profile factors, disease restrictions and correlation between taken diet and performed exercise. SHADE is dynamic and interactive in which diet and exercise suggestions are based on recent glucose level, taken diet and performed exercise. The suggestion mechanism needs concepts and information from various domains knowledge base, so a modular approach is used in which each domain (personal profile, food, disease and exercise) concepts and relationships are defined in a separate OWL based ontology and then these individual domain ontologies are combined into an integrated ontology to generate diet and exercise recommendations for diabetes patients. The recommendations are actually inferences by Pellet reasoner on the basis of predefined SWRL rules. The rest of the paper is organized as follow: Section 2 comprises of related work about various approaches towards food and exercise recommendation systems. Section 3 is about the mechanism of food and exercise recommendations. Section 4 describes modeling of various ontologies and their integration. Our prototype implementation is discussed in section 5. Section 6 demonstrates the results of our prototype application and contribution of our work is presented in section 7. Finally, conclusion and future work is described in section 8. II. RELATED WORK A diabetes patient needs personalized diet recommendation according to his/her preferences, calorie needs and nutrients

requirements. A personalized food and exercise recommendation system can assist a diabetic in controlling blood glucose level and to prevent the complications. Following articles have made efforts in this direction. Lee et. al. [8][9] describes fuzzy set definition and models fuzzy ontology for personal profile and food domain. Their proposed Intelligent Diet Recommendation Agent (IDRA) uses fuzzy set based inference mechanism to recommend diet menu for dinner. IDRA analyzes whole day taken meal, find out additional/lacking calories and nutrition for each taken meal and adjusts extra/deficient nutrition and calories in recommended dinner diet menu. IDRA lacks in defining relationships between food items and diabetes, such as which food item is recommended or not for diabetes. Further, exercise recommendations are completely ignored. Villarreal et al. [10] proposed an ontology based framework for patients mobile monitoring. As a case study, to model diabetes patients monitoring food, disease and personal profile ontologies are combined. The personal profile ontology lacks in presenting user’s preferences about food. Exercise is an important factor with respect to diabetes patient monitoring, which was ignored. The framework has not paid any attention towards rules, reasoning and inferences. J. Cantais et al. [11] developed an OWL based food ontology using Protégé for its integration and knowledge sharing with their PIPS project. The ontology was enriched with various classes, properties and food item instances along with their nutrition values. The development process and relevant challenges and difficulties have discussed. Reasoner was used just used to validate the ontology while ignoring its main functionality of rules and inferences. After the food ontology development, further work is not in our knowledge. Jong et al. [12] expressed the idea of a personalized diet recommendation system for heart patients. Predefined XML format is used for knowledge sharing and integration mechanism among different domains. Implementation details, diet recommendation output and results have not discussed. Gergely et al. [13] and Abdus Salam et al. [14] proposed a case based approach for diet recommendation. A ripple down rule mechanism is presented with a graph, where a condition or attribute value is checked at each node and depending upon the value of the attribute predefined subsequent action is recommended. User preferences about food items and food recommendations in menu format are not considered in [13]. An automated procedure for the recommendation is adapted in [14]. When an expert manually defines action(s) for a particular case, the system stores the case along with the action in knowledge base, so that it can be taken or advised automatically when similar case happens next time. Different domains such as personal profile, food and disease etc. knowledge is considered as a single system without using any knowledge sharing mechanism among different domains. En-Yu Lin et al. [15] proposed an intelligent consultation and recommendation framework to facilitate nutritionists using Hadoop cloud computing environment. Data mining approach is used to develop personalized and comprehensive nutrition programs for customers on weekly basis. A single

database contains different domains data as a single system. It lacks in standard data sharing and integration mechanisms among various domains. Clustering approach is used [16] to group food items based on their different nutrient values definition. Each cluster itself contains normal, limited and avoidable foods from diabetes point of view. Patients can check a food item and if it is forbidden, system suggests alternative food items having approximate similar nutrient values. [17, 18] used OWL based ontology to model person and exercise domains with the help of Protégé to generate health advices and exercise recommendations. Rules and reasoning based approach generates exercise recommendations as inferences. Both person and exercise domains knowledge are mixed and defined in a single ontology. Diet recommendations were out of scope. The uniqueness of our work is that we have used semantic OWL based ontologies to model different domains and then combine knowledge as an integrated ontology. Further, rules and reasoning approach is deployed that generates personalized food and exercise recommendations as inferences. The mechanism for generating the recommendations is described in next section. III. SHADE RECOMMENDATIONS SHADE suggests a user personalized and preferred diet menu and exercise recommendations. The recommendations are validated by nutritionist or dietician and if any enhancements are needed, rules are modified with coordination of domain and system experts. The recommendations may be generated on user demand or automatic system triggers. A. Diet Recommendations Diet menu is generated in two phases; Fist phase estimates the amount of calories and nutrients that must be satisfies for the meal menu, to be generated. Second phase extracts the food items which are users preferred, having no negative expected impact on disease situation and then create various alternative menus from food items such that menu total calories and nutrition values are in acceptable range. First of all, user’s body mass index (BMI) is calculated so that if he/she is over or under weight, daily intake calories can be adjusted. Basal metabolic rate is estimated using Harris- Benedict [19, 20] formula, as depicted by equation 2 and 3, and then multiply it with user’s corresponding physical activity factor to estimate daily calorie need. We have enhanced the BMR equation by replacing actual weight with ideal weight. Body Mass Index (BMI) = actual weight (in Kg) / (height (in meters))2 (1) BMR (Male) = 66.47 + (13.75 * w) + (5.0 * h) – (6.75 * a) BMR (Female) = 655.09 + (9.56 * w) + (1.85 * h) – (4.67 * a)

(2) (3)

where ‘h’ is height in cm, ‘a’ is age in years and ‘w’ is the ideal weight in kg, which can be estimated using following equation. w=

TABLE I. CATEGORIES AND ACTIVITY FACTOR FOR PERSON EXERCISE Person’s life style Activity Daily extra calories for exercise Factor for exercise Nil (BMR is resting energy Little or no exercise 1.2 and 1.2 factor consume in little daily routine work) Light exercise (1.375 – 1.2) * BMR 1.375 (1–3 days per week) = 0.175 * BMR Moderate exercise (1.55 – 1.2) * BMR 1.55 (3–5 days per week) = 0.35 * BMR Heavy exercise (1.725 – 1.2) * BMR 1.725 (6–7 days per week) = 0.525 * BMR Very heavy exercise (1.9 * – 1.2) * BMR (twice per day, extra 1.9 = 0.7 * BMR heavy workouts)

needed calories to be burned using equation 5 and second step filters user’s preferred and disease allowed activities and then calculate duration for recommended activities using equation 6. planCaloriesToBurn = Daily Calories burned by exercise (Column 3 of table 1) + Calories adjusted for overweight + extraIntakeCalories – caloriesAlreadyBurned (5)

The calories burned during exercise/activities can be calculated using following formula [23]: Calories burned (CB) = duration (in minutes) *MET * 3.5 * weight (in Kg) / 200

Daily calorie estimate = BMR * Activity Factor

(4)

where Activity Factor can be determined from Table I. Daily calories common distribution for each meal type is breakfast: 20%, lunch: 25%, dinner: 25%, snacks: 30% (10 mid breakfast + 10 evening snacks + 10 bed tea). The key important nutrients that affect blood sugar level are carbohydrates, proteins and fats [21].The recommended daily intake nutrients measure is carbohydrates 50%-60%, protein 10%-20% and fat 25-35% of daily calories [22]. Such distribution varies on individual basis and defined by nutritionist in the beginning. Based on diet and blood glucose level (BGL) monitoring, the distribution can be updated. Diet and exercise are strongly correlated. A diabetes patient with doing more exercise can enjoy more calories and in absence of exercise he/she need to be careful about calories intake, so we have devised an algorithm that dynamically allow change in number of calories for meal on a specific day, depending upon the physical activities of patient on that day and number of calories already taken in prior meal on that day. In phase 2, list of food items is retrieved and then filtered for disease prohibited food items and user’s non-preferred food items. The remaining food items are allowable, preferred and recommended food items. SHADE recommends a number of alternative meal menus such that each menu comprises of a list of food items having sum of nutrients and calorie values remains in allowable range and the diet is and balanced. B. Exercise Recommendations Exercise recommendations needs to consider person’s willingness for exercise, age, weight, extra taken calories, preferences and disease profile. Exercise recommendations mechanism also follows two steps. First step estimates the SWRL/OWL Rules for Person Food Domain Ontology

Inferred Knowledge

Inferred Knowledge

Reasoner

Person Knowledge base

Diet and exercise suggestion for diabetics needs to share and combine information from various domains. These domains include foods, exercise, personal information, diabetes domain, physical activity or exercise domain and then integration of these domains. Ontology based system is best suitable for various reasons including knowledge base creation, separation of knowledge with rules, reasoning capability and knowledge sharing situation. We used modular approach and developed separate OWL based ontology for each domain using Protégé 4.1. Individual ontologies works independently without effecting change in one ontology into another. Further, due to distribution of each domain knowledge separately, ontology size remains manageable and controllable. SHADE main architecture is shown in Fig. 1. A. The Domain Ontologies Personal Profile Domain Ontology: Person domain ontology is shown in Fig. 2. Personal health profile ontology models two types of user health profile variables: basic input parameters and derivable information from input parameters. Basic input parameters includes gender, age, weight, height, physical life style, calories and nutrients (carbohydrates, proteins, fats) percentage for each meal type (breakfast, lunch, dinner, snacks) while derivable information includes BMI, ideal weight, BMR, per day estimated energy need, number of calories and nutrients (in grams) for each meal type etc. Classes are defined in description logic based semantic class definition and SWRL rules. For example, a ‘Young’ is defined SWRL/OWL Rules for Exercise

Inferred Knowledge

Reasoner

Food Taken

Vital Signs

Exercise Done

Inter Domain Integrated Knowledge base

Suggestion

User’s Data Reasoner

SWRL/OWL Rules for Diabetes Disease Domain Ontology

Reasoner

Exercise Knowledge base

Preferences

Integrated Ontology

Figure 1: SHADE Architecture

IV. ONTOLOGY MODELING AND INTEGRATION

Exercise Domain Ontology

Food Knowledge base

SWRL/OWL Rules for Inter domain

Inferred Knowledge

where MET is metabolic equivalent, and it is per minute consumption of oxygen in the body at the state of sitting at rest. The value 3.5 is amount of oxygen per kg body weight per minute [24]. There is a specific value of MET for each combination of activity type and intensity level. For example, if aerobic exercise is done with low intensity then its MET value is 5.04. Such MET values can be seen on net, such as [25, 26].

SWRL/OWL Rules for food

Person Domain Ontology

(6)

Inferred Knowledge

Reasoner

Disease Knowledge base

Input

Output

User Interface

user

Food/Diet Suggestion Exercise Suggestion Reminders

Monitoring Nutrionist

/ Dietician

Figure 2: Personal Profile Domain Ontology

Figure 4: Inference of all food items having 7 grams of Proteins

as a person having age value in the range of 18 and 30.

Fig. 4 also shows instance ‘Apple’ is inferred as Fruit, RawFoodItem and MediumCalorieFoodItem on the basis of defined SWRL rules and semantic class definition. In order to find out various alternative food items having the same nutrient values, we have defined following SWRL rule. As an example of this rule, Fig. 4 presents inference of all food items having 7 grams of proteins.

Person and (hasAge some decimal[> 18 , OverWeight(?p)

SWRL based rules are defined to estimate the derivable information using the mathematical equations. For example, BMI is calculated using following SWRL rule: Person(?p), hasHeightInCms(?p, ?h), hasWeightInKgs(?p, ?w), divide(?bmi, ?w, ?hms), divide(?hm, ?h, 100), multiply(?hms, ?hm, ?hm) -> hasBMI(?p, ?bmi)

Food Domain Ontology: Food ontology has designed and then defined in OWL using protégé. Objects properties are defined for relationships between concepts such as Food_Item, Food_group, Nutrients and MealMenu etc. The ontology has enriched with numerous instances of food items along with portion sizes, different nutrients and calorie values. Standard nutrients values of food items are extracted from USDA [27] and [28]. Using rules and reasoning approach, food items are automatically categorized on the basis of different food groups, cooking methods and various nutrient values clustering. Fig. 3 shows modeling of food ontology in protégé.

Figure 3: Food Domain Ontology

FoodItem(?f), hasNutrientMeasure(?f, ?nm), hasNutrients(?nm, ?n) -> isFoundIn(?n, ?f)

Nutrient and calorie clustering such as LowCarbFoodItem, HighFatFoodItem and LowProteinFoodItem are formed by defining SWRL rules which enables reasoner automatically classify a food item into a particular cluster of a nutrient. For example, following function defines the clustering of food items on the basis of amount of calories per serving. If n is number of calories in per serving of a food item then

The partial presentation of above function in the form of SWRL rule is as follow: FoodItem(?f), hasNutrientMeasure(?f, ?nm), hasNutrients(?nm, ?n), Calorie(?n), hasNutrientValue(?n, ?nv), lessThanOrEqual(?nv, 50) -> LowCalorieFoodItem(?f)

Figure 5: Integrated Ontology

Disease Domain Ontology: It is a general disease domain ontology which is designed to model disease knowledge base, related concepts and their relationships. We have used OWL semantic class definition and SWRL based rules for diagnosis and conformation of disease, depending on vital sign values. It includes hierarchical representation of various diseases and relationships of different diseases with particular vital signs. Clinical guidelines of disease diagnosis and verification are transformed into SWRL rules and automatically inferences are generated as disease diagnoses depending on vital sign values. For example, following rule shows diagnosis of diabetes patient, if he/she has glucose level value higher than 180. GlucoseLevel(?gl), Person(?p), hasVitalSign(?p, ?gl), hasVitalSignValue(?gl, ?gv), greaterThan(?gv, 180) -> DiabeticPatient(?p)

Exercise Domain Ontology: Exercise or physical activity domain ontology is designed and developed to model its domain knowledge. Hierarchy of various physical activities is structured, object type properties are defined and instances of various activities are generated with their possible intensities and metabolic equivalent (MET) values. Activities can be classified on the basis of MET values as inferences and respective function for activity intensity ‘AI’ and transformed SWRL rule is depicted as below:

Activity(?a), hasMET(?a, ?m), greaterThan(?m, 4), lessThanOrEqual(?m, 8) -> ModerateIntensityActivity(?a),hasIntensityValue(?a,"Moderate")

On the basis of rules, reasoning engine infers intensity wise classification of the activity. Some context variables also introduced to make it context-aware such that location of exercise, time of exercise etc. B. The Integrated Ontology Protégé allows reusability of knowledge base by importing existing ontologies and facilitates to further extend the knowledge base. The reason for choosing pellet is it is the only open source reasoner which is protégé as well as Jena compatible. A comparison and survey of various reasoners is presented by [29]. In our implementation, personal profile, food, disease and physical activity domain ontologies are imported into an integrated ontology and defines properties and SWRL rules over their concepts to generate diet and exercise suggestions as an inference. For example, concepts between disease and food ontology are defined as there are restrictedly used, allowed and prohibited food items in case of a particular disease. Furthermore, user’s preferred food items and exercises are defined. The integrated domain ontology is shown in Fig.5 which shows recommended food items and exercises as inferences. Following rule suggests low calorie food items for diabetics DiabeticPatient(?p), LowCaloriePreferredFoodItems(?f), hasPreferredFood(?p, ?f)  hasFoodSuggestion(?p, ?f)

Another example of exercise recommendation is suggested exercise for old person is low intensity exercise Old_Age(?p), LowIntensityActivity(?a), hasExerciseSuggestion(?p, ?a)

hasPreferredExercise(?p,

?a)



V. PROTOTYPE IMPLEMENTATION In order to work as a prototype application, ontology modeling and integration work as defined in previous section can also be done programmatically using Jena and Pellet API, both are written in Java. Jena allows to dynamically generating ontology or loading an existing ontology from a file. Jena APIs also facilitates in importing multiple ontologies and combine them into an integrated model so that to enable queries over integrated model. Existing domain ontologies which were modeled using protégé and were saved in separate files are reused by loading them into the application using Jena programming code. With the help of API, our prototype application enables queries over the integrated model to present the data on graphical user interface and to extend the model dynamically and programmatically by adding further tuples. Application user interface is developed in Java which enables to register new user, add/edit user’s health profile data, include new food items and their calories and nutrient values with portion size, insert new exercise along with MET values etc. The application allows user to enter data of taken meal, performed exercise and blood glucose level so that this data is processed and considered while next recommendation is generated. Diet and exercise recommendations interface is shown in Fig. 6 (a) and Fig. 6 (b). VI. RESULTS Suggested food items are actually inferences as a result of reasoning by pellet reasoning engine on the basis of predefined rules. Using suggested food items, various alternative menu options are generated, where each menu is a combination or list of suggested food items with the restriction of satisfying nutrients and calorie values constraints. Finally eliminate those combinations or menus which are invalid. Example of an invalid menu is a combination of chapati with rice. Fig. 6 (a) shows the recommended suggestion for diet whereas Fig. 6 (b) shows exercise suggestions with duration and intensity. The recommendations are as per user’s preferences and non-risky to the disease, so adaptable and usable. VII. CONTRIBUTION We have used first ever approach of OWL based ontology, semantic ontological domain concepts definitions and SWRL based rules and reasoning approach to generate diet and exercise suggestions. Previously, no reasoning and inference based technique has been used for diet and exercise assistance for diabetics. Furthermore, our implementation integrates knowledge from various domains for diet and exercise suggestions. No such consideration has been done previously. Our integration approach shows the real benefit of using ontologies i. e. integration and sharing of concepts from various domains, defines new relationships among these domains concepts and integrate them as a single system without affecting individual ontologies. Our generated food recommendations are in a menu format which makes our effort to be used in real and practical life.

[7]

[8]

[9]

[10]

[11] Figure 6 (a): Diet recommendations [12]

[13]

[14]

[15]

Figure 6 (b): Exercise recommendations

VIII. CONCLUSION AND FUTURE WORK Managing balanced diet and proper exercise by a diabetic is a complex task. Food and exercise recommendations for diabetics need to consider various knowledge base domains. In our current work we developed OWL based person, food, disease and exercise ontologies and then an integrated ontology combined these ontologies. Using semantic ontology based rules and reasoning approach food and exercise recommendations for diabetics are generated. SHADE evaluation is to be done by diabetes patients, endocrinologists and nutritionists for its validation and further improvements. Our future work includes incorporation of context domain such as location based services for diet and exercise recommendations. Moreover, social and persuasive aspect can be added. Finally, data mining approach can be used on user taken meals and performed exercise data so that new relationships can be explored among food items, exercises and blood glucose level.

[16]

[17]

[18]

[19]

[20] [21]

[22] [23] [24]

[25]

REFERENCES [1] [2]

[3] [4] [5] [6]

http://diabetespakistan.com/treatment/2013/05/08/diabetes-statistics-inpakistan/ Last accessed on 14 August 2014. Daily Nation, Nov 15, 2012. http://www.nation.com.pk/pakistan-newsnewspaper-daily-english-online/islamabad/15-Nov-2012/-diabetes-canavoid-limb-amputation-thru-preventive-steps retrieved on 14 Aug 2014 http://www.w3.org/standards/semanticweb/ accessed on 14 Aug 2014 http://www.w3.org/TR/owl-features/ Last accessed on 14 Aug 2014 http://www.w3.org/Submission/SWRL/ Last accessed on 14 Aug 2014 http://clarkparsia.com/pellet/ Last accessed on 14 Aug 2014

[26] [27] [28] [29]

Abdullah A.Aljumah, Mohammed Gulam Ahamad, Mohammad Khubeb Siddiqui “Application of data mining: Diabetes healthcare in young and old patients” Journal of King Saud University –Computer and Information Sciences (2013) 25,127–136 C. S. Lee , M. H. Wang, H. C. Li and W. H. Chen “Intelligent Ontological Agent for Diabetic Food Recommendation” 2008 IEEE International Conference on Fuzzy Systems (FUZZ 2008) C. S. Lee, M. H. Wang, H. Hagras “A Type-2 Fuzzy Ontology and Its Application to Personal Diabetic-Diet Recommendation” IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 18, NO. 2, APR 2010 V. Villarreal, R. Hervás, A. D. Fdez and J. Bravo “APPLYING ONTOLOGIES IN THE DEVELOPMENT OF PATIENT MOBILE MONITORING FRAMEWORK” 2nd International Conference on eHealth and Bioengineering - EHB 2009, 17-18th September, 2009, IaºiConstanþa, Romania J. Cantais, D. Dominguez, V. Gigante, L. Laera and V. Tamma “An example of food ontology for diabetes control”. in Working notes of the ISWC 2005 Workshop on Ontology Patterns for the Semantic Web. 2005. Galway, Ireland. Jong-Hun Kim, Jung-Hyun Lee, Jee-Song Park, Young-Ho Lee, KeeWook Rim “Design of Diet Recommendation System for Healthcare Service Based on User Information” 2009 Fourth International Conference on Computer Sciences and Convergence Information Technology Gergely Kov´asznai “Developing an Expert System for Diet Recommendation” 6th International Symposium on Applied Computational Intelligence and Informatics, May 10-21 2011, Timisoara, Romania Abdus Salam Khan, Achim Hoffmann “Building a case-based diet recommendation system without a knowledge engineer” Artificial Intelligence in Medicine 27 (2003) 155–179 En-Yu Lin, Don-Lin Yang, Ming-Chuan Hung “System Design of an Intelligent Nutrition Consultation and Recommendation Model” 2012 9th International Conference on Ubiquitous Intelligence and Computing and 9th International Conference on Autonomic and Trusted Computing Maiyaporn Phanich, Phathrajarin Pholkul, and Suphakant Phimoltares “Food Recommendation Systemrte Using Clustering Analysis for Diabetic Patients” S. Izumi, D. Kuriyama, G. Itabashi, A. Togashi, Y. Kato and K. Takahashi, “An Ontology-based Advice System for Health and Exercise,” Proc. the 10th IASTED International Conference on Internet and Multimedia Systems and Applications 535-029, pp.95-100, 2006. S. Izumi, D. Kuriyama, Y. Miura, N. Yasuda, R. Yotsukura, Y. Kato and K. Takahashi, “Design and Implementation of an Ontology-based Health Support System,” Technical Report of IEICE SS2006-82, pp. 19.24, 2007 J. Arthur Harris and Francis G. Benedict. “A Biometric Study of Human Basal Metabolism Proceedings of the National Academy of Sciences”. Vol. 4, No. 12 (December 1918): 370–373. J. Arthur Harris and Francis G. Benedict. “A Biometric Study of Basal Metabolism in Man” Washington, DC: Carnegie Institution, 1919. Huan-Chung LI, Wei-Min Ko “Automated Food Ontology Construction Mechanism for Diabetes Diet Care” Proceedings of the Sixth International Conference on Machine Learning and Cybernetics, Hong Kong, 19-22 August 2007 http://www.nutristrategy.com/ Last accessed on 14 Aug 2014 http://www.drgily.com/activity-calories-burned.php?ex_id=2030 Last accessed on 14 Aug 2014 M. JETTE. K. SIDNEY.* G. BLUMCHENt “Metabolic Equivalents (METS) in Exercise Testing, Exercise Prescription, and Evaluation of Functional Capacity” Compendium of Physical Activity: https://sites.google.com/site/compendiumofphysicalactivities/home Last accessed on 14 Aug 2014 http://juststand.org/portals/3/literature/compendium-of-physicalactivities.pdf Last accessed on 14 Aug 2014 http://ndb.nal.usda.gov/ndb/nutrients/index retrieved on 14 Aug 2014. http://www.myfitnesspal.com/food/search retrieved on 14 Aug 2014. Sunitha Abburu “A Survey on Ontology Reasoners and Comparison” International Journal of Computer Applications (0975 – 8887) Volume 57– No.17, November 2012