Chapter 4

RESULTS AND DISCUSSION

 

4.1       EXPERIMENT

Experiments on sample
data conducted in proposed DSS and results brought up in table 4.1. Random
samples collected from the data set to compare and examine the output accuracy
of DSS. There are total 20 random samples evaluated and summary obtained as
follow.

Sr.No.

Number of times pregnant

Plasma glucose

Diastolic blood pressure

Triceps skin fold

2-Hour serum insulin

BMI

Diabetes pedigree function

Age

Sample Class

DSS Output

1

10

139

80

0

0

27.1

1.441

57

0

The patient is predicted negative for diabetes.

2

11

143

94

33

146

36.6

0.254

51

1

The patient is predicted positive for diabetes.

3

7

133

84

0

0

40.2

0.696

37

0

The patient is predicted negative for diabetes.

4

2

100

66

20

90

32.9

0.867

28

1

The patient is predicted positive for diabetes.

5

15

136

70

32

110

37.1

0.153

43

1

The patient is predicted positive for diabetes.

6

1

81

72

18

40

26.6

0.283

24

0

The patient is predicted negative for diabetes.

7

1

117

88

24

145

34.5

0.403

40

1

The patient is predicted positive for diabetes.

8

2

125

60

20

140

33.8

0.088

31

0

The patient is predicted negative for diabetes.

9

2

106

64

35

119

30.5

1.400

34

0

The patient is predicted negative for diabetes.

10

6

134

70

23

130

35.4

0.542

29

1

The patient is predicted positive for diabetes.

11

6

87

80

0

0

23.0

0.084

32

0

The patient is predicted negative for diabetes.

12

30

107

62

13

48

22.9

0.678

23

1

The patient is predicted positive for diabetes.

13

0

140

65

26

130

42.6

0.431

24

1

The patient is predicted positive for diabetes.

14

0

101

76

0

0

35.7

0.198

26

0

The patient is predicted negative for diabetes.

15

10

161

68

23

132

25.5

0.326

47

1

The patient is predicted positive for diabetes.

16

12

106

80

0

0

23.6

137

44

0

The patient is predicted negative for diabetes.

17

1

95

60

18

58

23.9

0.260

22

0

The patient is predicted negative for diabetes.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

18

3

187

70

22

200

36.4

0.408

36

1

The patient is predicted positive for diabetes.

19

0

135

94

46

145

40.6

0.284

26

0

The patient is predicted negative for diabetes.

20

3

129

92

49

155

36.4

0.968

32

1

The patient is predicted positive for diabetes.

 

Table.
4.1 Summary of Experiments

There
are 49 rules generated for C4.5 classifier on given dataset to predict the new
record. To evaluate DSS performance we conducted an experiment on randomly choose
20 samples and compared the output with original values.

 

Fig.
4.1 Rules Snapshot

4.2       PROPOSED DSS TRAINING / LEARNING

In the proposed DSS
first interface is for DSS Learning / Training. User can select algorithm for
prediction from available two algorithms that are C4.5 and SVM. After the
selection of predictor user may click to “Start Learning”. Model is train and reply
back with a message “Successfully built the model. Provide prediction
record.”  With this, DSS also bring up
Dataset Characteristics where information about dataset is presented with No.
of Instances, No. of Attributes and Class Labels.

Fig.
4.2 Proposed DSS Training/Learning Interface

 

4.3       PROPOSED DSS PREDICTION

Prediction is done in
DSS Prediction panel on user interface. Fig 4.3 shows the prediction interface
of DSS. Provide New Patient Readings is panel designed for user to enter
patient reading to predict diabetes orientation. This interface is mapped with
dataset attributes. User need to click the “Predict” button for prediction of
new record.

Fig.
4.3 Proposed DSS Prediction Interface

 

4.4       PROPOSED DSS OUTPUT

Fig. 4.4 is
representation of Output panel. A descriptive output is generated by the system
about new record provided to DSS for prediction. The output of system can be “The patient is
predicted positive for diabetes.” Or
it can be
“The patient is predicted negative for diabetes.” Output
panel also contain summary that present detail accuracy with confusion matrix
values.

Fig.
4.4 Proposed DSS Output

 

4.4       PREDICTION WITH PROPOSED DSS

 A new sample record is provided to the DSS and
a snapshot taken for detail elaboration in below Fig. 4.4. Proposed DSS work
well and accordingly for new readings. Proposed DSS can work as a standalone in
any environment or can be integrate with any expert system or information
system.

 

Fig.
4.4 Proposed DSS Output

 

 

CONCLUSION AND FUTURE WORK

 

With
emerging growth Information Technology, there is great need to implement
intelligent system in medical domain. Organizations very keen to implement
Decision Support in Expert Medical System to provide better care to patients
and to facilitate clinicians for better diagnosis. Thus, a good Decision
Support System with design and complete implementation can play a pivotal role
in any medical practice. Practically implementation of conducted researches is
a big challenge for expert developers to integrate with working models. This
proposed DSS may work standalone or easily integrated with working expert
system. We may need to focus on how easily any proposed DSS can be integrated with
working WEBEHR, PMS or PHR for smooth functioning.

SUMMARY

 

Implementation
of Data Mining algorithms in common programming languages is a challenging work
for software developers. Therefore, a practical implementation of Data Mining
algorithms may facilitate software developers to integrate Decision Support
with software’s to enhance the worth and functionality. Proposed DSS designed
and implemented in such a way that could be easily integrated with any expert
system or electronic health records. This Proposed DSS implemented by using
Visual C# 2017 in Microsoft Visual Studio Community 2017 Version 15.4.1 with
Microsoft .NET Framework Version 4.6.01055.