AI in Healthcare: Machine Learning for Disease Diagnosis & Prediction

This essay examines the role of artificial intelligence in healthcare, focusing on machine learning applications for disease diagnosis and prediction. It analyzes two articles: one comparing multi-view and single-view machine learning for disease diagnosis in primary healthcare, and the other discussing machine learning for disease prediction. The analysis covers the methodologies, variables, and results of each study, highlighting the potential of AI to improve disease diagnosis, prediction, and overall healthcare outcomes. The essay concludes by emphasizing the increasing significance of AI in medicine and its potential to revolutionize patient care through efficient and accurate diagnostic tools.

Running head: ARTIFICIAL INTELLIGENCE AND HEALTH CARE
Artificial Intelligence and Health Care
Name of the Student
Name of the University
Author Note

1ARTIFICIAL INTELLIGENCE AND HEALTH CARE
Introduction
The importance of machines can be felt to a wide extent in the contemporary period. This
is so because the human population has been on the rise after the end of the two world wars.
Moreover, the leaders of the nation states emphasize on the provision of a quality standard of
living to the people. As a result, to meet the demands and wants of the people, machines have
been developed since they are the ones who can deliver efficient and timely services and goods
to the people. They are becoming more preferable than the functioning of the human beings
engaged in such industries because there is, no matter how insignificant, a scope of human error.
On the other hand, machines, if properly constructed and organized, will be effective in
executing its prefixed functioning in an efficient manner, without marginal or negligible error.
This essay seeks to analyze the importance of artificial intelligence through the medium of
machines in the different field of an individual’s life. For this purpose, two articles shall be
analyzed, one titled “Multi-view versus single-view machine learning for disease diagnosis in
primary healthcare” written by Labroski (2018) and the other written by Frandsen (2016) entitled
“Machine Learning for Disease Prediction”. The essay concludes with the predominating role
played by such machines and the rise in significance of the same in the human world.
Article 1: Multi-view versus single-view machine learning for disease
diagnosis in primary healthcare
Summary of the Article
In this article titled “Multi-view versus single-view machine learning for disease
diagnosis in primary healthcare” written by Labroski (2018) talks about the role of machines and
their ability to solve the problems that occur in the sphere of primary health care. There are two
kinds of machines that are used in this sector, such as multi view machines and single view

2ARTIFICIAL INTELLIGENCE AND HEALTH CARE
machines. These machines are considered to be helpful in trying to diagnose the disease and
predict the issue of the concerned patient based on the past historical medical records of the
individual.
Hypothesis statement of the article
The hypothesis of the article is that it tries to analyze the competence of the two machines
of predicting the issue of disease and sees which of the machines is capable of getting an optimal
solution for the disease faster and at a more efficient rate.
Variables studied
The variables studied in this article is the prevailing health of a concerned individual and
the efficiency of the machines, that is the single view machine and the multi view machine, in
being able to predict and diagnose the disease and give a competent solution in trying to
overcome this issue in an efficient manner. The factors which were considered in the studying of
this research were the economic and social environment, the physical environment and the
behaviors and characteristics of the concerned individuals.
Sample studied in the article
The sample studied in the article includes those people who are in need of primary health
care.
Method of data collection
The paradigm of the multi view machine learning was chosen as an approach that will be
able to improve and enhance the predictive performance of classifiers in situations where there is
availability of many data sources which are heterogeneous in nature. For the comparison of the
single view machine learning with the multi view machine learning, different experiments were

3ARTIFICIAL INTELLIGENCE AND HEALTH CARE
taken in to consideration. The sample surveyed were the people who had health conditions and
came to take the aid of primary healthcare centers. While under taking the research different
concepts of machine learning were touched upon such as the concept of ensemble learning,
dimensionality reduction based learning and stacked generalization.
Results of the article
The result shown in this article is that the paradigm of multi view stacked generalization
is quite powerful and can help to improve and enhance the predictive performance in a
supervised learning setting. The performance of the different models were evaluated using the F1
scores. Through this, the author has observed that there is an increase in the performance of 0.04
and a maximum increase of 0.114 F1 score points. The results also revealed that the functioning
of the multi view stacked ensemble learning is compatible as a noise reduction approach and
works extremely effectively in the situations where there is a prediction that significant amount
of noise will be produced. This can prove to be effective in the research projects where the
characteristics of such projects are not manually selected by the different domain experts.
The results produced by this article shows to its readers the extent to which artificial
intelligence has taken a dominant role within the lives of the individuals. The prediction of future
disease by the simple diagnosis of the past historical records of an individual is quite significant
and marvelous to imagine as it will be able to solve all the problems that are currently infecting
the human population. By taking a precautionary approach, the individuals can now choose to be
treated by likely diseases, thereby increasing their chances at surviving for a longer period of
time. As a consequence, it can be noted that this research done in the article “Multi-view versus
single-view machine learning for disease diagnosis in primary healthcare” is of significant

4ARTIFICIAL INTELLIGENCE AND HEALTH CARE
contribution to the progress of human kind and their perpetual existence with the help of
artificial intelligence and machines.
Article 2: Machine Learning for Disease Prediction
Summary of the Article
The author Frandsen (2016) of the article titled “Machine Learning for Disease
Prediction” is of the opinion that thousands of people, mainly from the country of the United
States of America. They suffer from diseases which often go undiagnosed or are diagnosed at
such a later stage and the implementation of any medical aid proves to be largely ineffective,
especially in the cases of Type II Diabetes and Chronic Kidney Disease. Such a situation has
occurred mainly due to to prevailing socio-economic issues of the country where majority of the
people are not able to afford even primary health care. However, doctors are of the opinion if
such diseases are caught at an earlier stage then the cost required for under taking preventive
health care course will be quite cheap as compared to the cost required for treating the disease at
a later stage. It will also improves the chances of survival of the individual.
Hypothesis statement of the article
In this article, the author has derived the development of certain algorithms which will
help to predict the occurrences of the various diseases by drawing inspiration from different
techniques and ideas in the sphere of machine learning.
Sample studied in the article
The sample size of this research has been carefully selected. The people in need of
medical attention are large in number in the United States of America. However, the population
who are able to afford it are relatively low. Thus, in order to get the real essence of the

5ARTIFICIAL INTELLIGENCE AND HEALTH CARE
population to be surveyed, the machine learning method is taken in to account, but carefully, so
that additional features are not included. The researchers in this article took the data set of two
health care centres belonging to two very different regions of the United States of America. One
of the characterizing feature of the data set is the insurance claims of the people during a
particular time frame. This is important because the sample population only includes such
individuals. Moreover, the insurance claims are a way of providing information to the
researchers if the concerned individuals were diagnosed with a particular target disease during
that period of time.
Method of data collection
The researcher gathered the data with the help of several models. Linear classifiers were
used as they have the advantage of being an easy model to use. Moreover, linear classifiers are
considered to be extremely efficient to aid in the solving of problem. Apart from classifiers
another method that was used in order to collect data was that of training. The different
techniques of machine learning are able to use their function through the use of observed data.
This process is termed as training and the entire set of data that is utilized is known as training
set. The aspect of training is important as it helps in the selection of a function or model that is
compatible with the training set. For the purpose of predicting diseases, two important
techniques which are used, according to the author of this article, are that of Cox proportional
hazards model and logistic regression.
The data gathering technique of training has one important problem of regularization. In
other words, the classifier training faces the problem of over fitting. This problem arises when
the training set fits well but it faces the problem of poor generalization. One of the most

6ARTIFICIAL INTELLIGENCE AND HEALTH CARE
important cause for this problem is that of noise. At times, the training set maybe too small or
generate a lot of noise and this in turn, cancels out the bigger data sets.
Results of the article
The result of the research in the article shows that the recurrent neural network trained on
the CCS representation of the ICD-9 code sequence is the strongest and the most effective. Such
an aspect is true for both CKD and DM. Moreover, the research shows that the recurrent neural
network has an advantage over the other techniques. The best performing recurrent neural
network classifiers has a size of about two hundred and fifty six. Such a result proves that further
gains is possible if the number of neurons is increased in that specific layer. To conclude, the
result of the experiment conducted in this article shows that the historical medical records of a
patient can be optimally harnessed to predict diseases. Moreover, the modelling of the data in a
complete sequential nature rather than combining them to their BOW representation can possible
lead to an improved predictive performance.
Comparison of both the articles
In comparing the article of “Multi-view versus single-view machine learning for disease
diagnosis in primary healthcare” written by Labroski (2018) and the other written by Frandsen
(2016) entitled “Machine Learning for Disease Prediction” it is found that both the articles talk
about the importance of artificial intelligence and their application in the different machines in
the sphere of health care. This is important to understand because the articles show the extent of
artificial intelligence within the lives of the individuals and how they dominate over some of the
most important functions of human beings. The treatment of diseases has always been considered
to be the most risky job as the very survival of a person is within the hands of another person.
This is quite risky in the sense that there is always a margin for human error. However, with the

7ARTIFICIAL INTELLIGENCE AND HEALTH CARE
introduction of artificial intelligence in the field of medicine has greatly revolutionized the field.
This is due to the fact that it is very rare for a machine to cause any error if they have been
properly organized. Another point of similarity between the two articles is that both of them
employs the technique of random forest in order to analyze the data that has been collected
through the medium of classification and regression.
However, a main point of difference among the two articles is that article entitled “Multi-
view versus single-view machine learning for disease diagnosis in primary healthcare” written by
Labroski (2018) talks about the implication of only two important machines, such as the multi
view machine learning and the single view machine learning. The research conducted in this
article sees the efficiency of the two machines and compares them with one another. In the result,
it was found out that the multi view machine learning is more capable and effective than the
single view machine learning. In the other article written by Frandsen (2016) entitled “Machine
Learning for Disease Prediction” no specific machines were selected. Instead, the article focused
on the importance of the different methods of data collection and how this data would be
analyzed. The results of this research showed that the recurrent neural network trained on the
CCS representation of the ICD-9 code sequence is the strongest and the most effective. From this
we can gather the different methods of making a machine effective was provided for in the
article, unlike the earlier article which focused on the significance and working of the two
machines.
Conclusion
In conclusion it is observed that the two articles talked about the importance of health
care and how to provide efficient measures which can help people to predict the diseases that
might occur. In this way, preventive measures can be under taken so that the survival chances of

8ARTIFICIAL INTELLIGENCE AND HEALTH CARE
the individual increases to a considerable extent. In the future course of action in the article titled
“Machine Learning for Disease Prediction” that was written by Frandsen (2016) the author was
of the opinion that simply being able to understand what diseases might occur is not enough.
There is a need to be able to treat such diseases. For this, there is a further need to be able to
develop the artificial intelligence to such a level that they can efficiently help people to survive.

9ARTIFICIAL INTELLIGENCE AND HEALTH CARE
References:
Frandsen, A. J. (2016). Machine Learning for Disease Prediction.
Labroski, A. (2018). Multi-view versus single-view machine learning for disease diagnosis in
primary healthcare.