Welcome to scholarfront.com Data mining
We are living in the data mining age. Provide an example on how data mining can turn a large collection of data into knowledge that can h Order original solution to this paper here.
Data mining
We are living in the data mining age. Provide an example on how data mining can turn a large collection of data into knowledge that can help meet a current global challenge in order to improve healthcare outcomes.
APA style. 250 words minimum. One reference.
IMAGE & SIGNAL PROCESSING
Schizophrenia Auxiliary Diagnosis System Based on Data
Mining Technology
Xiaohong Wang1 & Na Zhao1 & Peng Ouyang2 & Jiayi Lin3 & Jian Hu1
Received: 26 December 2018 /Accepted: 13 February 2019
# Springer Science+Business Media, LLC, part of Springer Nature 2019
Abstract
In order to use digital medical technology to develop and design an auxiliary diagnosis system for schizophrenia to assist doctors
at all levels to diagnose and predict the cure of patients, improve the accuracy of diagnosis of symptoms, find complications in
advance, and reduce the risk of disease, the application of Bayesian network in auxiliary diagnosis system of schizophrenia is
studied, and an auxiliary diagnosis system of schizophrenia is designed. Based on data mining technology, knowledge informa-
tion can be found from patient data and used to diagnose the nature of patients. The demand analysis of auxiliary diagnosis
system is briefly introduced, and an auxiliary diagnosis system for schizophrenia based on Bayesian network is designed.
Keywords Bayesian network . Auxiliary diagnosis system . Demand analysis . Functional design . Data mining
Introduction
Schizophrenia is a serious mental disease, and the incidence is
unexplained. Emotion, perception, thinking and behavior and
other aspects of the disorder and mental activities and other
symptoms are often related to syndrome in clinical [1]. The
incidence rate is 0.007% – 0.014%, often having onset in
young adults, and the clinical cure rate is low, which brings
a lot of burden to the patients and their families [2]. In life, the
general awareness of patients with schizophrenia are relatively
clear and intelligence is basically normal, but in the disease
process, the course generally has a long period of time, and the
disease will deteriorate fast. As a result, the cognitive function
of patients is damaged, and it will cause mental decline or
mental disability [3]. But some patients, after a scientific and
reasonable treatment, can achieve rehabilitation or basic
rehabilitation.
If digital medical technology can be used to develop a
schizophrenia auxiliary diagnosis system to assist doctors
at all levels to diagnose and predict schizophrenia pa-
tients, the accuracy of diagnosis can be improved, symp-
toms can be detected in advance and the risk of onset can
be reduced [4]. Digital intelligent medical technology is
the trend of the development of medical information tech-
nology. It solves the Binformation island^ phenomenon
among different medical institutions and realizes the shar-
ing of medical information among different institutions
[5]. With the continuous development of the scale of hos-
pital clinics, the number of patients in hospital clinics is
also increasing, the number of patrols per unit time is
increasing sharply, the patient information management
is also very complex and diversified, and the traditional
way is gradually difficult to meet the requirements of
patients on the level of service, seriously affecting the
operational efficiency of medical institutions and hinder-
ing the development of medical institutions [6].
Therefore, it is necessary to build a mobile patrol infor-
mation management and query platform for schizo-
phrenics. Wireless technology is an important method
and means to update the information and data of schizo-
phrenia patients to the central system quickly and effec-
tively, to solve various problems in the mobile patrol of
schizophrenia patients for medical institutions, and to
This article is part of the Topical Collection on Image & Signal
Processing
* Jian Hu
[emailprotected]
1 Department of Psychiatry, The First Affiliated Hospital of Harbin
Medical University, 23 Youzheng Street, Nangang District,
Harbin 150001, China
2 School of Management, Harbin Institute of Technology,
Harbin 150001, China
3 Beijing Electro-Mechanical Engineering Institute, Beijing 100074,
China
Journal of Medical Systems (2019) 43:125
https://doi.org/10.1007/s10916-019-1214-8
http://crossmark.crossref.org/dialog/?doi=10.1007/s10916-019-1214-8&domain=pdf
mailto:[emailprotected]
improve the service level and management efficiency of
departments [7].
The research content of schizophrenia auxiliary diag-
nosis system is to integrate and analyse individual genetic
background data, health data, disease-related molecular
biology data and drug clinical trial data, and to form a
network electronic health system technology and data
analysis system to study disease prediction, diagnosis,
treatment, and prevention digital medical knowledge anal-
ysis method and its integrated software [8]. From a bio-
logical and medical point of view, it is difficult for biol-
ogists to discover the effects of a single or several genes
on organisms and the relationship between them as a
whole by manipulating them. However, with the develop-
ment of technology, it is now possible to analyse personal
health indicators, medical records, drug reactions and oth-
er data [9]. At the same time, genetic information, protein
family tree information, genome-wide expression and
methylation information, as well as epigenetic information
can also be analysed. If biologically multi-dimensional
and multi-directional data can be fused organically, a pa-
tient can be described completely, thus achieving precise
medical purposes for schizophrenics [10].
Method
Demand analysis of auxiliary diagnosis system
for schizophrenia
Data requirement analysis based on Bayesian algo-
rithms: According to the data characteristic information
required, case report forms are designed and subjects
are selected in the research hospital centers. Finally,
316 schizophrenic patients are selected as data to verify
the model of auxiliary diagnosis system. Among them,
the selected patient data contains 237 dimension attri-
butes and 8 of 237 dimension attributes are category
attributes. 17 attributes have more vacancy values, va-
cancy rate is about 13%, 79 attributes have discrete data
values, and other attributes are continuous data values
[11]. In the latter study, the selected schizophrenic pa-
tients sample data will be used for learning and training
to obtain decision rules, and to explore whether the data
demand characteristics can better meet the needs of the
system, and then be used for clinical auxiliary
diagnosis.
316 patients with high-dimensional small sample data
are selected for two main purposes: to study the impact of
the attributes of clinical samples on their categories, and
to explore whether the patient data meet the needs of the
auxiliary diagnosis model to guide the diagnosis process;
to find a method of mining high-dimensional small
sample data. The main reason for the analysis of high-
dimensional small sample data is that in some cases or
in a short time only some data can be obtained, and
knowledge can be obtained from these data, so it is nec-
essary to study high-dimensional small sample data [12].
The most important thing here is how to ensure and im-
prove the accuracy of the results of this auxiliary diagno-
sis model.
By mining and analyzing the small sample data in clin-
ical diagnosis, we can no longer be limited by the number
of samples; the important condition attribute set obtained
can help doctors check only a few important items when
examining patients, which can not only reduce the diag-
nosis cost of patients, but also optimize the allocation of
medical resources. The results obtained after mining and
analysis can be applied to the auxiliary diagnosis system,
and then help or assist doctors to diagnose schizophrenia
of patients.
System function business process requirement analysis:
The main purpose of data pre-processing is to process the
data with redundancy, incompleteness, noise and high di-
mensionality that cannot directly use Bayesian network,
to provide simple, clean, accurate and normal data for the
auxiliary diagnosis system of schizophrenia, and to im-
prove the efficiency and accuracy of the auxiliary diagno-
sis system information processing of schizophrenia.
Therefore, the flow chart of data pre-processing that
the auxiliary diagnosis system should adopt is shown in
Fig. 1.
The auxiliary diagnosis system of schizophrenia based
on Bayesian network is composed of Bayesian network
structure and parameters. The Bayesian network can be
used to obtain the Bayesian network structure from the
patient sample data set through structural learning, then
to learn the parameters, and finally to obtain the parame-
ters of the Bayesian network. The construction process of
Bayesian network is basically the same, and its workflow
is shown in Fig. 2.
In fact, the process of auxiliary diagnosis of schizo-
phrenia by Bayesian network is to use the Bayesian net-
work has been built to calculate and analyse the newly
added patients data, and to judge the type of the input
patients data. As a result, reasoning diagnosis is actually
a problem of classification. The process of diagnosis
should first preprocess the original information of pa-
tients, standardize the patient records, and then calculate
the patient records using Bayesian network to get the di-
agnosis results. The diagnostic workflow is shown in
Fig. 3.
The update of the sample database mainly refers to adding
new patient data information to the sample database. The pro-
cess of updating is to manage the pre-processing of the patient
information that has been diagnosed and input it into the
125 Page 2 of 7 J Med Syst (2019) 43:125
sample database to obtain a new sample database. The
workflow is shown in Fig. 4.
Demand analysis of auxiliary diagnosis function model
based on Bayesian network. Naive Bayesian is an important
branch of Bayesian decision theory. Naive Bayesian hypoth-
esis requires that the value of an attribute affects a given class
independently of other attribute values and it is a supervised
learning method. Although this harsh restriction is often not
met in reality, naive Bayesian reasoning usually implements
attribute selection process first in data sets, which improves
the independence of attributes. Moreover, naive Bayesian rea-
soning can generate more complex non-linear decision-mak-
ing surfaces, and can fit fairly complex surfaces and achieve
great success.
Based on the improved naive Bayesian method of attribute
weighting, the predictive formula of the patients untreated
probability can be obtained according to the formula as fol-
lows:
P C2=Xj
P Xj=C2
P C2
P Xj=C1
P Xj=C2
P C2
1
This involves the estimation of the class conditional prob-
ability density. P(Xk|Cj) can be obtained from the training set
by fitting the class conditional probability density (that is, the
probability density function of Xk) of the characteristic attri-
bute component Xk in each grouping Cj. According to the
value type of attribute variable Xk, the estimation methods
of class conditional probability density are different.
When XK is a discrete numerical value, then:
P Kk=Cj
Njk
Ni
2
When it is a continuous numerical value, according to the
improved naive Bayesian model method mentioned above,
that is, the conditional probability density function fitting Xk
by the kernel density estimation method according to formula
Xk, P(Xk|Cj) is calculated as follows:
P Xk=Cj
1
nh
n
t1
K
XkXt
h
3
K(x) is called the kernel function, h is called the window
width of the kernel function, that is, if the larger h is chosen,
the deviation may be larger, and the estimated probability
density function will be smoother; if smaller, the estimated
probability density function will not be so smoother, but the
probability density curve and sample fitting will be relatively
better.
The Logistic regression method and the naive Bayesian
method before and after improvement are used to establish a
model to predict the cure probability (PHM) of schizophrenia
patients in the course of treatment. The model is applied to the
auxiliary diagnosis system. The resolution performance of the
three models on validating the data set of schizophrenia pa-
tients is shown in Fig. 5.
Among them, the area under ROC (Receiver Operating
Characteristic) curve of Logistic regression model is AUC
(Area under concentration-time curve) = 0.5142 0.1095,
standard naive Bayesian model is AUC = 0.5899 0.1063,
and improved naive Bayesian model is AUC = 0.7721
0.0865. The difference has statistical significance
(P
The post Data mining
We are living in the data mining age. Provide an example on how data mining can turn a large collection of data into knowledge that can h first appeared on Scholarfront Essay Writing Service. Order original solution to this paper here.
