**6. Conclusions**

The proposed system uses different Machine learning classifiers to recognize and categorize faults in Wireless Sensor networks. The dataset has four major classes, they are DoS, Probe, R2L, U2R which are further categorized. In this paper for the purpose of fault detection Random Forest (RF), Support Vector Machine (SVM), Stochastic

*Efficient Machine Learning Classifier for Fault Detection in Wireless Sensor Networks DOI: http://dx.doi.org/10.5772/intechopen.111462*


#### **Table 10.**

*Performance of SGD for varying fault rate.*

Gradient Descent (SGD), Multi-layer Perceptron (MLP) a classifiers are used to classify sensed data into faulty and non-faulty data Fault detection is a challenging task since wireless networks are placed in confined spaces. Machine learning classifiers are employed in this project because they are effective. The ML algorithms are trained using preprocessed data sets. One Hot Encoding is the method that is used to pre-process the data. Since in the data set few columns does not contain Numeric values. Recursive feature elimination is used to select the features that are applicable and which helps to find the specific attack. The system is put to the test on data set that were not seen during the training phase, some new attacks are introduced in the test data and the result show that the system is effective in identifying faults in the WSN. Since fault detection in the WSN can be challenging, due to harsh environment where the WSN are deployed makes them vulnerable to faults. Therefore machine learning is essential for fault detection since it is less time consuming, faster and also gives the good accuracy.
