SIMULATION RESULTS AND DISCUSSION

Table of Contents:

This section examines the classification performance obtained by the SMOTE-FSVM model using PIMA Indians Diabetes Dataset. It comprises a total of 768 instances with 8 attributes and 2 classes. A set of 34.90% of samples comes under positive class and the remaining 65.10% of samples come under negative class. The details are listed out in Table 7.1. Among the total number of 500 negative instances and 268 positive instances, the SMOTE model upsamples the data instances and convert 536 instances as positive. Thereby, the number of instances in the two class labels is effectively balanced.

Figure 7.2 analyzes the performance of the SMOTE-FSVM model in terms of precision. The figure portrayed that the DT model is found to be the ineffective performer which has attained a minimal precision of 81.40%. Next to that, the Logitboost model has outperformed the DT model by attaining a precision of 84.60%. Besides, the SVM model has reached to a certainly higher precision of 85.32%. In line with, the LR and FSVM models have led to near acceptable precision values of 88% and 87.46%, respectively. However, the presented SMOTE-FSVM model has achieved better performance by attaining maximum precision of 91.47%.

Figure 7.3 analyzes the performance of the SMOTE-FSVM model in terms of recall. The figure portrayed that the LogitBoost model is found to be the ineffective performer which has attained a minimal recall of 77.61%. Next to that, the DT model has outperformed the earlier model by attaining a recall of 79.02%. Besides, the LR model has reached to a certainly higher recall of 79.27%. In line with, the SVM and FSVM models have led to near acceptable recall values of 80.12 and 82.99%, respectively. However, the presented SMOTE- FSVM model has achieved better performance by attaining maximum recall of 89.63%.

Figure 7.4 analyzes the performance of the SMOTE-FSVM model in terms of accuracy. The figure portrayed that the DT model is found to be the ineffective performer which has attained a minimal accuracy of 73.82%. Next to that, the LogitBoost model has outperformed the DT model by attaining an accuracy of 74.08%. Besides, the LR model has reached to a certainly higher accuracy of 77.21%. In line with, the SVM and FSVM models have led to near acceptable accuracy values of 76.65% and 80.11%, respectively. However, the presented SMOTE-FSVM model has achieved better performance by attaining maximum accuracy of 85.52%.

Figure 7.5 analyzes the performance of the SMOTE-FSVM model in terms of F score. The figure portrayed that the DT model is found to be the ineffective performer which has attained a minimal F score of 80.19%. Next to that, the Logitboost model has outperformed

TABLE 7.1 Dataset Description

Description

Value

Number of samples

768

Number of features

8

Number of classes

2

% of positive samples

34.90%

% of negative samples

65.10%

Data source

[20]

Precision analysis of SMOTE-FSVM model

FIGURE 7.2 Precision analysis of SMOTE-FSVM model.

the DT model by attaining an F score of 80.95%. Besides, the LR model has reached to a certainly higher F score of 83.41%. In line with, the SVM and FSVM models have led to near acceptable F score values of 85.83% and 84.57%, respectively. However, the presented SMOTE-FSVM model has achieved better performance by attaining maximum F score of 90.45%.

Accuracy analysis of SMOTE-FSVM model

FIGURE 7.4 Accuracy analysis of SMOTE-FSVM model.

Table 7.2 provides the response time analysis of SMOTE-FSVM model over the compared models. The table values denoted that the LogitBoost model has required high computation time of 40 seconds and offered higher response time compared to other models. Besides, the LR and DT models have attained somewhat lower response times of 36 and 35 seconds, respectively. Followed by, the proposed SMOTE-FSVM model has achieved moderate response time of 32 seconds which is higher than previous method not than FSVM and SVM models. Fiowever, the FSVM and SVM models have required minimal response time of 28 and 25 seconds, respectively. The experimental validation indicated

TABLE 7.2 Response Time Analysis

Classifiers

Time (second)

SMOTE-FSVM

32

Fuzzy support vector machine

28

Support vector machine

25

Logistic regression

36

LogitBoost

40

Decision tree

35

that the SMOTE-FSVM model has attained superior classification results over the compared methods. Besides, the usage of SMOTE model has increased the classifier performance to the next level.

CONCLUSION

This chapter has developed a novel SMOTE-FSVM model for class imbalance problem in IoT- and cloud-based disease diagnosis. The medical data has been gathered from a diverse set of sources, namely IoT sensor data, medical records, and UCI repository. The patient’s data are acquired and transmitted to the cloud. Then, SMOTE-based upsampling process takes place to resolve the class imbalance problem. Finally, FSVM-based data classification task is carried out to determine the class labels and find the existence of the diseases. The classification performance of the SMOTE-FSVM model has been validated using PIMA Indians Diabetes dataset. The simulation outcome verified that the SMOTE-FSVM model has obtained maximum results with the precision of 91.47%, recall of 89.63%, accuracy of 85.52%, and F score of 90.45%.

REFERENCES

  • 1. Yin Y, Zeng Y, Chen X, Fan Y (2016) The internet of things in healthcare: an overview. / Ind Inflntegr 1:3-13.
  • 2. Chen M, Zhang Y, Qiu M, Guizani N, Hao Y (2018) SPHA: smart personal health advisor based on deep analytics. IEEE Commun 56(3):164-169.
  • 3. Hu L et al. (2016) Internet of Things cloud: architecture and implementation. IEEE Commun Mag 54(12 Supp):32-39.
  • 4. Chen M et al. (2018) Edge-CoCaCo: Toward joint optimization of computation, caching, and communication on edge cloud. IEEE Wireless Commun 25(3):21-27.
  • 5. Muhammad G, Rahman SKMM, Alelaiwi A, Alamri A (2017) Smart health solution integrating IoT and cloud: a case study of voice pathology monitoring. IEEE Commun Mag 55(1):69—73.
  • 6. Hossain MS, Muhammad G (2016) Cloud-assisted industrial Internet of Things (IIoT) enabled framework for health monitoring. Comput Netw 101:192-202.
  • 7. Gope P, Hwang T (2016) BSN-Care: a secure IoT-based modern healthcare. IEEE Sens J 16(5):1368-1376.
  • 8. Priyanka K, Tripathi NK (2015) A real-time health monitoring system for remote cardiac patients using smartphone and wearable sensors. Int J Telemed Appl. doi: 10.1155/2015/373474
  • 9. Gelogo YE, Hwang HJ, Kim H (2015) Internet of Things (IoT) framework for u-healthcare system. Int J Smart Home 9(11):323—330.
  • 10. Hussain A, Wenbi R, Lopes A, Nadher M, Mudhish M (2015) Health and emergency-care platform for the elderly and disabled people in the Smart City. J Syst Softw 110:253-263.
  • 11. Catarinucci L, De Donno D, Mainetti L, Palano L, Patrono L, Stefanizzi ML, Tarricone L (2015) An IoT-aware architecture for smart healthcare systems. IEEE Internet Things 2(6):515-526.
  • 12. Xu B, Da Xu L, Member S, Cai H, Xie C, Hu J, Bu F (2014) Ubiquitous data accessing method in IoT-based information system for emergency medical services. IEEE Trans Ind Inf 10(2):1578—1586.
  • 13. Box IM, Yang G, Xie L, Mantysalo M, Zhou X, Pang Z, Da Xu L, Member S (2014) A health- IoT platform based on the integration of intelligent packaging, unobtrusive bio-sensor and intelligent medicine box. IEEE Trans Ind Inf 10(4):2180—2191.
  • 14. Maia P, Batista T, Cavalcante E, Baffa A, Delicato FC, Pires PF, Zomaya A (2014) A web platform for interconnecting body sensors and improving health care. Proced Comput Sci 40:135-142
  • 15. La HJ (2016) A conceptual framework for trajectory-based medical analytics with IoT contexts. J Comput Syst Sci 82(4):610-626.
  • 16. Hajihashemi Z, Popescu M, Member S (2016) A multidimensional time-series similarity measure with applications to eldercare monitoring. IEEE J Biomed Health Inf20(3):953-962.
  • 17. Liu B, Li J, Chen C, Tan W, Member S, Chen Q, Zhou M (2015) Efficient motif discovery for large-scale time series in healthcare. IEEE Trans Ind Inf ll(3):583-590.
  • 18. Chawla NV, Bowyer KW, Hall LO, Kegelmeyer WP (2002) SMOTE: synthetic marginal over-sampling technique. J Artiflnt Res 16:321e357.
  • 19. Lin CF, Wang SD. (2002) Fuzzy support vector machines. IEEE Trans Neural Netw 13(2):464-471.
  • 20. https://www.kaggle.com/uciml/pima-indians-diabetes-database.
 
Source
< Prev   CONTENTS   Source   Next >