Intelligent Signal Analysis Using Case-Based Reasoning for Decision Support in Stress Management - Computational Intelligence in Healthcare: Advanced Methodologies - page 174

The performance of the system in terms of accuracy has been compared with
experts in the domain, where the main goal is to see how close the system could work
compared to an expert. The evaluation in this chapter, considers several test data sets
to illustrate the overall system performance compare to an expert.
The initial case base comprised of 39 FT measurements as reference cases from 24
subjects. Seven woman and seventeen men between the age ranges of 24 to 51
participated in this study. The cases, in their conditional or problem description part,
contain a vector with the extracted features and the solution part comprised of
expert's defined classification as diagnosis. The levels of stress are classified by the
expert into five classes ranging from 1 to 5 where 1=VeryStressed, 2=Stressed,
3=Normal/Stable, 4=Relaxed and 5=VeryRelaxed.
The performance of the three matching algorithms: modified distance, similarity
matrix, and fuzzy matching are evaluated where the test is done for the three test
groups for three different query cases. Both in ranking and in similarity performance,
fuzzy similarity matching algorithm shows better results than the other algorithms
(i.e. distance function and similarity matrix) when compared with the expert's
opinion [71].
For this experiment, 5 test groups consisting of various numbers of cases have been
created (i.e. TG-A=5, TG-B=7, TG-C=9, TG-D=11 and TG-E=14) where cases are
selected randomly and all the cases are classified by the expert. These formulated test
cases were then used in the classification process by the CBR system using the fuzzy
similarity matching algorithm. The main goal of this experimental work was to see
how the CBR classification system supported the diagnosis of stress compared to
an expert.

Table 5. Experimental results for the test groups

Test
Group

Number of
Cases

Goodness-of-fit
(R 2 )

Absolute mean
Difference

TG-A 5 0.94 0.20

TG-B 7 0.92 0.14

TG-C 9 0.67 0.33

TG-D 11 0.78 0.30

TG-E 14 0.83 0.28

Average 9.2 0.83 0.25

The results of the experiment for each test group are illustrated in Table 5. As can
be seen from Table 5, the first two columns present the name and the number of the
cases for each test group. The classification of the cases in each group by the CBR
system is then compared with the expert's classification. Goodness-of fit (R2) [45]
and absolute mean difference (error) by the system for these five groups have been
calculated and presented in Table 5. R2 values of all the sets are 0.94, 0.92, 0.67, 0.78
and 0.83; absolute mean differences of the five sets are 0.20, 0.14, 0.33, 0.30 and
0.28; so the average R2 and error values of these sets are 0.83 and 0.25, respectively.