Page 283 - Big Data Analytics for Intelligent Healthcare Management

11.6 RESULTS ANALYSIS 277

11.6 RESULTS ANALYSIS
The projected approach was executed in MATLAB and all the perspectives were assessed on eight
numbers of biomedical datasets from the UCI repository [38] and Kentridge repository [39]. The data-
sets are considered as same as in [40].

11.6.1 EVALUATION METRICS
In this study, to evaluate the potential of the anticipated technique, two evaluation metrics (objective
function and accurateness) were measured. Accuracy (Eq. 11.9) is the proportion of the number of true
positives and negatives for the whole number of instances.
TP + TN

Accuracy ¼ (11.9)
TP + TN + FP + FN
Here, the terms TP, TN, FP, FN stand for true positive, true negative, false positive, and false negative.
In addition to that, the standard of clusters produced through the algorithm can be specified by clus-
tering accuracy (Eq. 11.10).
No: of correctly sampled data
Clustering Accuracy ¼ 100% (11.10)
Total no: of sample
The clustering accuracy was calculated for every dataset and the proposed method was compared with
the outcomes.

11.6.2 EXPERIMENTAL RESULTS
Experimental analysis was performed using six techniques comprising the anticipated method. Out-
comes were attained for all the datasets. As the proposed method is based on clustering and the aspect
of performance is accuracy, the essential standard technique of clustering known as FCM and several
additional techniques (SVM, Naive Bayes, Back propagation neural network (BPNN), and decision
tree) were measured for the experimental assessment. Every technique was trialed for 50 independent
runs and the consequences were reproduced. As the composite approach was utilized to discover the
preeminent clusters, i.e., it is obliging to evade from fixed at local minima that is predominantly occurs
in FCM. However due to the initialization is throughout preeminent cluster hubs, the projected advance
was well-organized when compared to FCM. The alterations in objective functions in every iteration
were directly proportional to the number of iterative loops in any algorithm. Here the intention is to
address the issues related to the number of iterations and objective function. It was obvious that the
efficiency of the projected technique was somewhat superior to FCM.
In Table 11.1, the average accuracy of all the methods is provided and the end result is that the
anticipated FCM-KA method is better than many other techniques. In the Thyroid and Dermatology
datasets, it is obvious that the recommended method has the highest accuracy of 96.81% and 98.42%
respectively. The superiority of the proposed method is clearly represented in in terms of accuracy. The
projected method was compared with several works on similar datasets (Fig. 11.2). As shown in four
other modern studies, the projected FCM-KA system showed better precision. Moreover, the outcome