. classified as descriptive model or predictive model.
June 5, 2019
. Data Mining Model for Classification
The objective of a data mining model is to create an understandable structure by taking the data set into account. Based on the model, the system behavior can be identified. Usually, the data mining model may be classified as descriptive model or predictive model. In the design of HIF detection method, predictive data mining model is preferred because of the requirement of the work i.e. Classification.
We Will Write a Custom Essay Specifically For You For Only $13.90/page!
There are several data mining models reported 16-18, However, DT model is chosen because of its transparency, efficiency and popularity. The proposed method utilizes open source data mining package ‘R’ for creating the DT model 19. For HIF detection, DT model is generated by considering the feature set into account.There are four features set training the DT against the target output of 1 for HIF and 0 for non-HIF. From these feature data set, 70% used for training and 30% for testing. The data mining generated DT model with three features set (Standard deviation, Skewness and Kurtosis) for HIF detection shown in Fig.5. As a result, these features have more discriminating capability in the classification of HIF from non-HIF.
IV. RESULTS AND DISCUSSIONS
The behavior of phase current with HIF inception time of 0.065 sec is shown in Fig. 6. The high frequency transients due to HIF are clearly seen in wavelet level D1 to D3. The response of current due to non-linear load switching at a time of 0.065 sec is shown in Fig.7. The behavior of non-linear load shows that, it has high peak value compared to HIF, which can be seen from the wavelet level D1 to D3. Another non-HIF event considered in the work is capacitance switching, the decomposed signal after this event inception time of 0.065 sec is shown in Fig.8. It is very clear from the wavelet level D2 to D4, there is no significant change in oscillation few cycles after event inception. After calculating features from the signal, subsequently apply the same to DT model to take final decision on the classification of HIF and non-HIF.
The proposed method is evaluated through the following three parameters:
1. Dependability: Predicted HIF against total HIF conditions.
2. Security: Predicted Non-HIF against total non-HIF conditions.
3. Accuracy: Actual predicted against the total number of conditions considered.
Testing was carried out against 278 cases comprises 243 HIF and 35 non-HIF cases. The confusion matrix of the DT model given in Table 3 infers that all the HIF are correctly classified. However, two non-HIF cases are misdirected as HIF and this false tripping due to the huge variation of non-linear load added to the system. Table 4 presents the performance indices of the proposed method. It is very clear that the method has excellent detection rate i.e. all HIF cases are detected even there was a huge amount of non-linear load variation.
Moreover, power system always susceptible to noise, it is necessary to investigate the proposed method with varying Signal to Noise Ratio (SNR). Table 5 compares the performance under noisy atmosphere. It is observed that the performance is same as in the normal case for 10 dB of SNR on the signal. On the other hand, for the 20 dB of SNR, one of the performance index termed as dependability (HIF detection rate) of the method holds good. However, other two indices slightly reduced. Additional boost in SNR, reduces all the performance indices. Hence, the proposed method is suitable for SNR of 20 or less.
Finally, Table 6 shows the comparison of proposed work with previously published paper. There is considerable improvement shown by the proposed method. The performance an index such as dependability and accuracy is good. Though all HIF cases classified correctly, small quantity of test cases considered in 6 and more than 10% of false tripping in 4. The proposed scheme is better in all the three performance indices.