Probability of a Device Failure using Support Vector Machine by comparing with Random Forest Algorithm to improve the accuracy
Degala Lokesh, J. Femila Roseline · 2023
Aim: The main purpose of this study is to compare the effectiveness of two methods for predicting a device's failure: the Innovative Support Vector Machine (SVM) and the Random Forest (RF). Materials and Methods: From the Kaggle dataset, 800 samples of device failures were collected. These samples were split into two groups: 560 for training (70%) and 240 for testing (30%). To determine the performance of the SVM algorithm, accuracy, precision, and specificity values were calculated.Results:Based on the overall performance analysis of independent samples t-test on the two groups, the SVM algorithm achieved accuracy, precision, and specificity of 86.6%, 96.20%, and 83.5%, respectively, compared to 78.40%, 77.68%, and 95.60% for the RF algorithm. These models were significant (p 0.05), and G power was found to be 0.8.Conclusion: In this study, the SVM algorithm outperforms the RF algorithm in detecting probability device failure.