Fig 6 shows the performances of different approaches using accuracy,
specificity, sensitivity and error rate parameters in a graphical
manner. It is observed that PSO-ANN approached obtains higher accuracy,
sensitivity, specificity rate and low error rate for dengue disease.
While, DT approach shows poor performance among all other approaches. It
is stated that significant differences occur between the performances of
the above mentioned approaches and concluded that PSO-ANN based
diagnostic model is more capable to detect and diagnosis of dengue
disease in earlier stage. Based on the experimental study, the following
point can be highlighted.
-
The higher accuracy rate can be achieved using PSO-ANN approach to
diagnosis of dengue patients. It can be possible by optimizing weight
(w) and bias (b) parameters of ANN using PSO technique. From the PSO
algorithm, initially optimized value of the weight and bias parameter
is determined and after that the optimized values pass in ANN
parameters.
-
It is also observed that slight variations are occurring between the
performances of NB and PSO approaches especially for sensitivity and
specificity parameters. It is found that the average rank of both
techniques is same i.e. 3.5. This indicates that both of the
techniques have similar performance for diagnosis of dengue disease.
But, the AUC performance measure indicates that NB approach is better
than PSO due to higher AUC rate.
-
To validate the diagnostic accuracy, AUC performance measure is also
used. This performance measure can be used to validate the diagnostic
accuracy obtained classification approach. The result of this
performance measure also confirms that the PSO-ANN based diagnostic
model is an effective and efficient model for earlier diagnosis and
detection of dengue affected patients.
[CHART]
Fig 6: shows the performance comparison of accuracy, sensitivity,
specificity and error rate of different approaches