FTVETI
School of Graduated Studies
Department of ICT
Thesis Proposal:
Efficient Predictive model to discover patient cause of
visit using Data mining approach
on Decision Trees(DT) and Naive Bayes (NB) Algorism
BY: ABDEREZAK KAMIL
EMERGENCY Department is the frontline for hospital to face patients in
1. INTRODUCTION
1.1 Background of the study
emergencies.
In recent years, hospital’s emergency service unit has been frustrated with a
lot of problems, which include, large number of patients’-(overcrowding),
delayed patient treatment, long processing time, and high costs.
Predicting causes of accidents are one of the most interesting and
challenging tasks in which to develop data mining applications.
In recent years new research avenues such as knowledge discovery in
databases (KDD), which includes data mining techniques, has become a
popular research tool
CONT…
Predictive data mining in clinical medicine enables to derive models that
use patient information to support specific clinical decisions.
Data mining models can be applied for building decision-making
procedures such as diagnosis, and treatment planning, which once evaluated
and verified.
Even though the application of data mining to medical data analysis has
been relatively limited until recently, the term data mining has been
increasingly used in the medical literature over the past few years.
1.2 Statement of the Problem
Emergency department are confronted to exceptional events due to seasonal
epidemics (in winter: influenza, colds, bronchiolitis, etc…in summer:
trauma) health crisis … these event can cause strain situation when
disequilibrium between the care load flow(demand) and care production
capacity(supply) is observed.
The management of patient flow in such situation is one of the most
important problem managed by ED managers. To handle this influx of
patients, EDs require significant human and material resources, but this are
limited .
CONT…
Generally there are a lot of problems and challenges regarding emergency
medical service in Ethiopia which is due to many factors such as shortage
of health professionals, facilities, resources and discovery of non-trivial
knowledge from the collected data.
The problems of previous research were not only related to the small
proportion of data set used but also the data analysis was conducted simple
statistical techniques. To this end, this work attempts to explore and answer
the following research questions:
Cont …
Therefore; this thesis addresses the following research
questions:
Which types of emergency cause of accidents are most
significant?
Which type of emergency case is at greatest risk of mortality?
What is the potential of data mining technique to predict major
cause of attending emergency medical services ?
1.3 OBJECTIVES
1.3.1 General Objective
The general objective of this study is to explore the potential of data mining technique
for predict causes of accident based on emergency medical data.
1.3.2 Specific Objectives
The specific objectives of this study are:
To explore the extent of problem of emergency medical service pertaining to big data
To collect and extract the dataset required for the problem domain.
To preprocess the raw data into a suitable dataset for model building.
To selected to be used for building the models can apply to the scenario under
investigation
To build data mining models on the preprocessed dataset
To analyze and interpret the results of the models.
To extract interesting patterns which are relevant to the problem domain
Evaluating the model built.
Report results and make recommendations
1.4 RESEARCH METHODOLOGY
The general purpose of the research is to build a model using
Data Mining Approach for Predicting Cause of Accident using
Emergency medical data based on the datasets collected from
the hospital emergency department.
1.4.1 Research Design
The overall research design follows hybrid model of
discovering hidden knowledge from the St. Paul hospital
emergency department through data mining technique. The
process model adopted to undertake this research is the hybrid
one due to the reason that this model describes each of the
knowledge discovery process steps in a better way and it is
flexible since it has a feedback mechanism in more steps than
the CRISP-DM.
The hybrid model has six steps that are : understanding of the
problem, understanding of the data , preparation of the data , data
mining , Evaluation of the discovered knowledge and use of the
discovered knowledge. The research is designed based on steps of
this process model
1.5 Scope and limitation of the research
This study mainly focuses on classification of St. Paul hospital In
the emergency department using data mining techniques were examined to
develop interesting classified and to extract meaningful patterns based on
their services common attribute data which were gathered from St. Paul
hospital.
In order to assess level of impact of each factor it is better to use different
data source which holds attributes in parallel with influencing factors. Here,
it is limited to develop the model using only St. Paul hospital in
the emergency department and doesn’t include other services in the
hospital.
1.6 Significance of the Study
The research was targeted to explore the benefit of
data mining technology for exploring significant
knowledge from the emergency medical dataset which
can be used to generate new or unseen knowledge that
can improve the current status of the emergency medical
health care by building applicable model and identifying
dominant factors related to the cause of visit.
The result of the study can also be used the predictive
model, as a frame work for keeping emergency cause of
visit in medical health care.
Moreover, the output of the study might be used as a
benchmark for other researchers who have interest in
such related issue or area
1.7 Work plan(Time Table)
Table 1 shows the work break down for the successful
completion of this work. It describes each activity
corresponding with the time length it will take to
accomplish it.
1.8 Reference
1. X. K. Zhou, “Discussion on the features of high emergency resources consumers by data
mining technology ”, National Taiwan University, Taiwan, 2004.
2. Fromm, R., McCallum,W. ,Niziol, C., Babcock, J., Gueler, A. & Levine ,R.( 1993). Critical
care in the emergency department: a time-based study. Crit. Care Med., 21, 970-976.
3. Huang, H. et al. “Business rule extraction from legacy code”, Proceedings of 20thInternational
Conference on Computer Software and Applications, IEEECOMPSAC’96, 1996, pp.162-167
4. Bellazzi, R. &Zupan, B. (2008). Predictive data mining in clinical medicine: current issues and
guidelines. International Journal of Medical Informatics, 77,81–97.
5. Cios Krzysztof J., PedryczWiltod.,Swiniarski Roman W. Kurgan Lukasz A. Data Mining: A
knowledge Discovery approach. New York: Springer-Verlag Science Business Media LLC;
2007Review of ensembles of multi-label classifiers: Models, experimental study and prospects.
Information Fusion 44: 33-45 (2018)
6. Witten I. H. and Frank E. 2005. Data Mining: Practical Machine Learning Tools and
Techniques, 2rd ed. Amsterdam: Morgan Kaufmann
7. [ Ethiopian Red Cross Society. First Aid/Community Based Healthand Road Safety [Internet].
Available from: <[Link] [accessed
20.06.2012].
Thank you