Essay:

Essay details:

  • Subject area(s): Engineering
  • Price: Free download
  • Published on: 7th September 2019
  • File format: Text
  • Number of pages: 2

Text preview of this essay:

This page is a preview - download the full version of this essay above.

Applications of Recommender Systems in Healthcare: A Review

Shivaprasad K.P

Department of Information Systems, Cleveland State University

1. ABSTRACT:

2. INTRODUCTION:

In our every day life, we use strategies fairly most of the time. For example, once we select a restaurant, we may ask friends about their interests and take their ideas into consideration. When we want to select a most important general practitioner or a health center, we may just pay attention to the reviews made through different users online or the suggestions from acquaintances. When we purchase merchandise by way of a web-based retailer, it is very fundamental for us to read stories from other shoppers to have information about their quality. In Facebook and other social networking websites, there mostly appear recommendations about people you may be aware of (Fang 2014).

Nonetheless information load and inappropriate understanding are important obstacles for drawing conclusion on the personal health status and taking proper actions (Chiu 2011). Confronted with a large quantity of clinical understanding on different channels (e.g., information sites, web forums, and many others.) customers more commonly get lost or consider unsure when investigating on their possess. Moreover, a manifold and heterogeneous medical vocabulary poses a further barrier for laymen. Consequently, increased personalized delivery of scientific content can aid users in finding central expertise. Scientific expertise on hand for patient-oriented determination making has accelerated greatly but is normally scattered across unique websites(Müller, Hanbury et al. 2012).

Growing health information and changes in information seeking behavior may also be discovered world wide. In keeping with latest experiences 81% of U.S. Adults use the web and 59% say they have got regarded on-line for healthcare understanding concerning ailments, diagnoses and one-of-a-kind remedies. Such effects impact the patients-healthcare professional relationship as informed patients raise questions or discuss remedy choices accordingly, patients are inclined to emerge as energetic contributors within the selection-making method. This change in the way of thinking is mostly referred to as patient empowerment.

3. BACKGROUND AND MOTIVATION:

4. RECOMMENDER SYSTEMS OVERVIEW:

Recommender system can be defined in various ways. It is operated by providing recommendations as an input from the user which is aggregated and directed to proper recipient. Based on prior record Recommender system can guess the items of user interest. Additional variant of this system shows a subjective nature of recommendations which gives personalized recommendations to different users with respect to personal interest of any item among the large pool of items collection.

The usage of recommender system in healthcare is growing with the passage of time. The provision of web connection allows firms and customers to provide and approach healthcare related data on-line. The patients are becoming mature in accessing health related information on-line. Recommender method usage has enabled users to entry understanding more adequately. The Wikipedia articles could be a very good source of understanding as it allows for users to go looking articles in so much better way than other engines like google given that of its structured advantage base. The relatedness may also be expanded additional by computing relevance with the aid of graph (Huang, Liu et al. 2012).

5. TECHNIQUES OF RECOMMENDER SYSTEMS:

Classification of Recommender systems is as follows: content-based, knowledge-based, collaborative, demographic-based and hybrid. These are based on background data and inputs. The Techniques of Recommendation is shown in Fig 1. Client(user) is the input source which feeds the input to the recommender system. Recommender systems gives the outputs in the different forms. One the basis of the users interests the recommendation systems are assigned. The complete information with the background data is processed before the process is started. (Musial2009).

Fig 1. Recommendation Techniques Classification

5.1. Collaborative Recommender Systems: Collaborative approach in general is structuring a model from a user's former behavior and alike decisions made by other users. It involves of information filtering using methods combining with various agents, etc., Collaborative filtering applications typically involve huge data sets. These techniques have been applied to various kinds of data including: sensing and monitoring data, like in environmental sensing; financial data, like financial service organizations that combine various sources; or in web applications and e-commerce in which the user data is focused. etc., (Kamran and Javed 2015).

5.2. Content-based Recommender Systems: On the basis of item details and users profile, these recommender systems are used. In this system, to define the items used words and to specify the item type liked by the user, profile of user is formed. The items are recommended by algorithms which are alike to those liked by a user in the past (Pazzani and Billsus 2007). In certain, different items are matched with items formerly valued by the user and recommend the items that are matched properly. This approach has its origins in information filtering search and information retrieval.

5.3. Demographic Recommender Systems: With respect to the user’s personal data Demographic recommender systems categorize the users and recommend items on the basis of demographic classes. It provides recommendation to the user by uniting the ratings of all users from a specific demographic nook. This has been made obvious from research that results of a study, conducted for a particular population, cannot be used to draw conclusion for another population if user sample differs too much. Demographics and user features drastically the recommendations. For an instance, a recommender system planned for tourism classifies the tourists according to their demographic information and recommend places on the basis of demographic classes. It considers that the similar category tourists have similar preferences(Trewin 2000, Beel, Langer et al. 2013). The recommendations produced by using demographic recommendation strategy are not very perfect.

5.4. Knowledge-based Recommender Systems:

The aim of all adapted recommender systems is to figure the items of interest for a particular user. Content based algorithms perform this task on the basis of interests of a user represented as text. Collaborative filtering algorithms perform this task on the basis of behavior of active user and other similar users.  This recommender systems uses the knowledge about user preference, item properties and criteria of recommendation(Trewin 2000). A huge quantity of data about user’s habits is vital to generate recommendations precisely. A sufficient quantity of rated items is essential for methods to be suitable. The knowledge-based recommender techniques don't suffer from this “ramp-up” problem as they are impartial on consumer ranks. The effectivity of knowledge-based recommender techniques is raised by this independency.

5.5. Utility-based Recommender Systems:

Based on the estimation of the utility of each item for the user, Utility-based recommender systems recommendations are made. Utility-based recommender techniques cause multi-attribute utility function built on rates of items suggested by the user to define predilections of user, and use the utility function to calculate item utility for user(Feng 2015). Utility-based strategies have not some problems of collaborative filtering recommender method, akin to high dimensions and cold-start. Nevertheless, utility-based recommender methods face eliciting utility perform for each user, which need first-rate burden of interaction. A couple of utility-based recommender systems had been developed centered on consumer score for gadgets they are situated on MAUT (Multi-Attribute Utility Theory) and needs consumer effort. So, modeling utility function with little consumer effort is an imperative quandary in designing utility-situated recommender techniques. Table.1 depicts the techniques and their functions with pros and cons.

APPROACHES

Collaborative

Content-based

Demographic

Knowledge-based

Utility-based

Functions

Aggregative ratings or recommendations of objects

Objects defined by their associated features

Categorize users based on personal attributes

Functional knowledge:  How a detailed item meets a unique need.

Make ideas  on the basis of computation of the each utility.

Functions

Recognize user similarities based on the ratings

Gain knowledge of consumer’s profile interests on the basis of the points in the objects the user has rated

Demographic classes recommendations

Can reason in regards to the association amongst a necessity and a viable advice

Employ constraint satisfaction techniques to locate the best match

Functions

Generate new references based on inter-user relationships

Long-term models, updated as more evidence about user preferences is observed

No long-term models

No long-term models

No long-term generalizations about users

Strengths

• Cross-genre niches can be identified

• No need of Domain knowledge

• Improvement of quality over time which is adaptive

• No need of Domain knowledge

• Improvement of quality over time which is adaptive

• Cross-genre niches can be identified

• No need of Domain knowledge

• Improvement of quality over time which is adaptive

• Ramp-up is not necessary

• Changes with the preference changes

• Non-product characteristics can be included.

• Mapping can be done from users to products

• Ramp-up is not necessary

• Changes with the preference changes

• Non-product

characteristics can be included

Weaknesses

• New user and item build-up problem

• Problem of “Grey Sheep”

• Large historical datasets quality dependency

• Plasticity problem versus Stability

• New user build-up problem

• Large historical datasets quality dependent

• Plasticity problem versus Stability

• Large historical datasets quality dependence

• Plasticity problem versus Stability

• Problem of “Grey Sheep”

• New user build-up problem

• Demographic information must be gathered

• Static suggestion ability

• Knowledge engineering is necessary.

• Utility function input must be given by the user.

• Static suggestion ability

Table 1. Techniques of Recommender systems, their functions with pros and cons.

6. APPLICATIONS OF RECOMMENDER SYSTEMS IN HEALTHCARE:

Researchers have indicated that combined healthcare information techniques are fitting an important part of the modern day healthcare systems. Such programs have developed to an integrated corporation-broad procedure. In exact, such programs are viewed as a type of enterprise information systems addressing healthcare enterprise sector desires. As a part of efforts, nursing care plan recommender systems can furnish clinical selection aid, nursing education, clinical satisfactory control, and function a complement to current apply instructional materials(Duan, Street et al. 2011). In this review paper, several applications for the healthcare recommender systems are discussed.

6.1. Personalized Wellness Therapy Recommendation Systems:

Increasing costs and dangers in health care have moved the preference of people from health remedy to disorder avoidance. This disorder avoidance is called as wellness. In recent years, the web has turn out to be a widespread situation for health-conscious users to seek for health-related knowledge and options. As the user community becomes more health conscious, service improvement is required to help users find applicable adapted wellness solutions. Because of fast progress in the wellness market, users value convenient access to wellness facilities. Thus,

the wellness industry should develop its Internet services in order to deliver better and more convenient customer service. The progress of a wellness recommender system progress would help users locate and adapt suitable personalized wellness therapy treatments bases on their discrete requirements. The accessibility and quality of wellness information delivery on the Internet is improved by these techniques. The wellness recommendation task is done using an Artificial Intelligence technique of hybrid case-based reasoning (HCBR). HCBR resolves user’s existing wellness problems by operating solutions from alike cases in the past. Wellness consultants use reliable wellness knowledge to recommend explanations for sample wellness cases generated across an online consultation form. Thus, the proposed model can be included into wellness websites to allow users to search for proper personalized wellness therapy treatment based on their health condition.

 The cause of using HCBR in this is to take experts knowledge and expertise alongside with formerly determined wellness cases in a case database. Users show their wellness concerns, which are resolved by locating an identical case in the case database (CBR). If the case database contains no adequately similar cases, the system will recommend a suitable healthcare solution by using pre-determined standard rules (Lim, Husain et al. 2013).

6.2. Personalized Health Education Recommender Systems:

The usage of computer systems in health education began greater than a decade in the past, more commonly for tailoring health academic assets. These days, among the laptop-tailoring well being education techniques are utilizing the web for providing one of a kind forms of well being schooling. Almost always, these systems are designed for exact sickness, with a predefined library of educational assets. These methods don't take abilities of the increasing quantity of academic resources on hand on the net. One of the causes is that the high availability of content material is making it extra complex to find the relevant one. The main issue of data overload has been addressed for many years within the recommender methods field (Fernandez-Luque, Karlsen et al. 2009). Computer-Tailoring Health Education Systems (CTHES) is an Expert Systems that automates many tasks achieved by the health educators.

CTHES

Recommender Systems

Health Education RS

Knowledge of medical expert for problem in health

Collaborative-based: on the basis of users who has similar problems

• Requires Critical users

• Knowledge lacking about users who are new

Approaches of RS to collect contents that are not considered by CTHES

Human experts dependent that makes tough for number of resources.

Content-based: on the basis of ratings

• Results are independent of other users.

• Information is required about items and ratings of users

• Better specialties of recommendations

Internet sources could be added.

6.3.  iDoctor: Professionalized and Personalized medical recommendations on the basis of hybrid matrix factorization:

iDoctor offers the user with professionalized and professionalized doctor recommendation with mining user emotion and preference from user rating and reviews about doctors Zhang, Y., et al. (2016). In specific, it has the modules and specific architecture also. Sentiment analysis module computes user emotional offset from user textual contents. Subject(Topic) modeling module is to excerpt the distribution of user preferences and doctor features. Hybrid Matrix Factorization(HMF) module, that is combined with feature distributions obtained by LDA (Latent Dirichlet Allocation) for valuation prediction(Luo, Thomas et al. 2012)

 Sentiment evaluation module computes consumer emotional offset from consumer reports textual content. Subject modeling module is to excerpt the distribution of user preferences and healthcare professional aspects. HMF module is built-in with two feature distributions obtained by means of LDA.

6.4. A Clinical Recommender System using data mining methods:

The combined medical information systems are the primary part of the today’s healthcare systems. These systems have developed to an integrated organization-wide system. In specific, these systems are regarded as a sort of enterprise information systems or Enterprise Resource Planning system for healthcare enterprise sector needs. As part of efforts, nursing care plan recommender systems can provide clinical nursing education, clinical quality control, decision support and function a complement to follow the guidelines(Duan, Street et al. 2011). Its implementation outdoes that of certainty, its assessable in real time. This Clinical Recommender system is a match to expert systems and usual practice rules, and can be very beneficial education in nursing and quality control in healthcare. It has the ability to connect systems, users, nursing care process in a way that allows nursing care to become more associated and responsive (Erol et al. 2010). Clinical Recommender system gives the services by paying attention on taking users information of their tasks which can be considered as recommendation (Wang et al.  2010).

6.5. mHealth Recommender Systems using Case Based Reasoning for Smoking Cessation:

 In these days many life-threatening diseases people are facing like cancers also related to cardiac disorders. This is all because people are addicted to tobacco and smoking. A unique smoking intervention plan in mHealth which uses a Case Based Recommender system with the help of mobile phones is brought in the recent days. It generates modified motivational messages on the basis of patient profile and same is delivered through mobile phones (Ghorai, Saha et al. 2013). This system introduces the smoking cessation intervention program which motivates the patients who have subscribed and they can quit smoking. The smoker’s behavioral data is used in the program and message list is sent via text messages. It uses Hybrid case based recommender systems (HCBR). Graded evaluation strategy is used by HCBR in sending the messages to users (Pattaraintakorn, Zaverucha et al. 2007). From the case database it locates similar cases by considering patients age and sex. Intensity function is used for the patterns of users based on the wellness target. Then it pacifies to moderate intensity messages. If this intervention fails, then it follows up with message of high intensity. Next it sets the user for automated voice call with a message as user has to quit smoking. Thus, it’s the first intervention plan to use mobile with case-based recommender system technique to motivate patients with messages.

6.6. GenNet Health-Care Social Network: Recommender System:

Several recommender system solutions have been found in recent days. Recommender systems in the social networking sites uses to recommend users. GenNet is a collaborative environment social network that helps in promoting the patients health if they have any disorder related to genes(de Magalhaes, Souza et al. 2013). This GenNet Health-Care social network recommender system improves the user communications and also the knowledge in medical related aspects(Arias, Vilas et al. 2012). As the friend numbers increases the information exchange also increases. And this improves in collaboration and communication within the patients and doctors with the medical information of the patients as well(Song, Dillon et al. 2011). In Social Networking Sites, researchers have used recommender systems to recommend the users with different user profiles for new connections. It uses friend-of friend algorithm. It has 2 scenarios that shows according to the ranking. Algorithm doesn’t show recommendations if there is no inputs (friends).

6.7. Medical Recommender Systems based on Evidence:

Evidence-based Medical Recommender System is a decision-making process by user assistance. This technology is used in medical field like in medical care and diagnosis. These systems use intelligent methods and they contain historical data which can be considered as medical knowledge and is used for Telemedicine (Sodsee and Komkhao 2013). In this, the recommendation is based on evidence by implying physician’s artificial knowledge. Physicians use EBM (Evidence-based Medicine) (Smith, Tong et al. 2006) with patient’s information on the basis of population. Evidence is grouped into two categories formal and informal where formal evidence is reports i.e., medical researches. And informal evidence is the opinions of experts. This recommender system has several algorithms like neural networks, fuzzy logic, support vector machine also data mining techniques that uses hybrid filtering techniques. For the history of patients collaborative filtering method is used; for experimental studies content-based technique is used. Systems can be accessed from anywhere by the physicians. Based on the patients physical examination the physicians will receive the recommendations and physicians in turn consult the patients. With all these data collection and cooperated evidences with the algorithms the strength of medical recommender systems can be understood. This effective Telemedicine is experimented in Thailand.

Healthcare RS Applications

Techniques used

References

Personalized Wellness Therapy

Hybrid case-based Recommender System: takes experts knowledge and expertise alongside with formerly determined wellness cases in a case database

(Lim, Husain et al. 2013)

iDoctor

Hybrid Matrix Factorization (HMF) is better than Basic Matrix Factorization (BMF). Confidence and Strength of Recommendations

(Luo, Thomas et al. 2012)

Clinical RS using Data mining techniques

It is a match to expert systems. gives the services by paying attention on taking users information of their tasks which can be considered as recommendation

(Duan, Street et al. 2011)

mHealth RS using case-based reasoning

Hybrid case-based Recommender systems (HCBR). Motivational messages to the patients. Graded evaluation strategy is used by HCBR in sending the messages to users

(Ghorai, Saha et al. 2013)

GenNet Health-Care Social Network RS

It is a Collaborative environment social network that helps in promoting the patients health if they have any disorder related to genes

(de Magalhaes, Souza et al. 2013)

Evidence-based Medical Recommender Systems

It uses collaborative filtering technique also algorithms like neural networks, fuzzy logic, support vector machine also data mining techniques that uses hybrid filtering techniques

(Sodsee and Komkhao 2013)

Table 2. Applications summarized for Healthcare Systems

...(download the rest of the essay above)

About this essay:

This essay was submitted to us by a student in order to help you with your studies.

If you use part of this page in your own work, you need to provide a citation, as follows:

Essay Sauce, . Available from:< https://www.essaysauce.com/essays/engineering/2016-5-6-1462507660.php > [Accessed 17.10.19].