The proposed iris segmentation's performance evaluation is comprehensively illustrated in this section. More promising results are seen in the model's outcomes. The presented system is accurate along with simple to execute which is beneficial for the proposed scheme. Analogizing the proposed system with prevailing research occurs in MATLAB (version 15), along with the proposed alcoholic drinker methodology is executed. The machine configuration is depicted as,
Processor: Intel core i7,CPU Speed: 3.20 GHz,OS: Windows 7,RAM: 4GB.
4.2 Performance Analysis
Regarding segmentation accuracy, sensitivity, specificity, precision, recall, f-measure, False Positive Rate (FPR), False Discovery Rate (FDR), Positive Predictive Value (PPV), Negative Predictive Value (NPV), together with Matthews correlation coefficient (MCC), the proposed MHCT’s performance is tested against the prevailing approaches like active contour, K-means clustering (K-means), Fuzzy C-Means Clustering (FCM), along with Particle Swarm Optimization (PSO)in this analysis section. Table 1 depicted the proposed approach’s analysis outcomes.
Table 1
Demonstrate the performance of the proposed MHCT with the existing methodologies
Metrics | Proposed MCHT | Active contour | K-means | FCM | PSO |
Accuracy | 0.99 | 0.84 | 0.55 | 0.60 | 0.51 |
Sensitivity | 0.99 | 0.84 | 0.61 | 0.59 | 0.58 |
Specificity | 0.92 | 0.91 | 0.33 | 0.50 | 0.02 |
Precision | 0.99 | 0.51 | 0.04 | 0.17 | 0.01 |
Recall | 0.92 | 0.92 | 0.33 | 0.50 | 0.02 |
F-measure | 0.96 | 0.59 | 0.07 | 0.25 | 0.01 |
FPR | 0.001 | 0.159 | 0.391 | 0.415 | 0.424 |
FDR | 0.003 | 0.486 | 0.957 | 0.827 | 0.995 |
PPV | 0.99 | 0.51 | 0.04 | 0.17 | 0.01 |
NPV | 0.99 | 0.08 | 0.85 | 0.95 | 0.82 |
MCC | 0.95 | 0.59 | -0.11 | 0.12 | -0.25 |
The proposed MCHT's performance with the prevailing active contour, K-means, FCM, and PSO algorithm are depicted in Table 1. By utilizing abovementioned metrics, the performance comparison is executed. Here, a lower performance was exhibited by the prevailing PSO than the proposed MCHT. Also, a high-level performance was offered by the proposed methodology than the prevailing active contour, K-means, and FCM clustering for all metrics. Consequently, the outcome illustrated that a high-level performance was attained by the proposed MCHT than the prevailing models. This analysis could be graphically represented below figure,
Regarding accuracy, sensitivity, together with specificity, the proposed MHCT's performance analogized with the prevailing active contour, K-means, FCM, and PSO are depicted in the above figure. Accuracy will bias towards sensitivity if it is high, or, towards specificity if it is high. Accuracy will also be high if both are high along with accuracy will be low if both are low. But, accuracy varies with specificity without pondering sensitivity in a case where the count of excellent candidates is low and the count of poor performers is high. Similarly, accuracy differs from sensitivity without pondering specificity in the case where the number of excellent candidates is high and the number of poor performers is low. Here, accuracy, sensitivity, and specificity value of 0.99%, 0.99%, and 0.92% were offered by the proposed MCHT, which is higher than the prevailing schemes. Hence, a better performance was provided by the proposed method than the prevailing schemes.
Regarding precision, recall, along with f-measure, the proposed MCHT’s performance was analogized with the prevailing active contour, K-means clustering, FCM, and PSO in the above figure. The number of positive class predictions that actually fit in to the positive class is quantified by precision. The number of positive class predictions made out of all positive examples in the dataset is quantified by Recall. A single score was provided by F-Measure that balanced both the precision and recall concerned in one number. Here, a precision, recall, along with f-measure of 0.99%, 0.92%, and 0.96% were attained by the proposed MCHT. Whereas the prevailing active contour, K-means, FCM and PSO have the precision of 0.51%, 0.04%, 0.17 5, and 0.01% and the recall of 0.92%, 0.33%, 0.50%, and 0.02% and the f-measure value of 0.59%, 0.07%, 0.25%, and 0.01%. Thus, the outcome suggests that a high-level performance was confirmed by the proposed scheme than the prevailing schemes.
With regards to the MCC metric, the above figure exhibits the proposed MCHT’sperformance with the prevailing active contour, K-means, FCM, and PSO. True and false positives and negatives were pondered by the coefficient. It is normally observed as a balanced measure that can be utilized even for very distinct size classes. A correlation coefficient betwixt the observed along with predicted binary classifications is the feature named MCC that returns a value betwixt − 1 and + 1. A perfect prediction was represented by a coefficient of + 1, nothing better than random prediction is signified by 0, along with total disagreement betwixt prediction and observation was indicated as − 1. MCC of 0.95% was attained by the proposed MCHT whereas the prevailing active contour, k-means, FCM, and PSO obtained a low value of 0.59%, -0.11%, 0.12%, and − 0.25%. Thus, a better performance was attained by the proposed method than the prevailing schemes.
Regarding PPV and NPV metrics, the proposed MCHT’s performance is analogized with the prevailing active contour, K-means, FCM, and PSO methods in the above figure. The system is regarded as a good system if it has the best PPV and NPV value. PPV and NPV value of 0.99% was offered by the proposed MCHT which is higher than the active contour, K-means, FCM, and PSO having PPV of 0.51%, 0.04%, 0.17%, and 0.01%, and NPV value of 0.08%, 0.85%, 0.95%, and 0.82%. Thus, the outcome revealed that the prevailing schemes were outperformed by the proposed MCHT.