Academia.eduAcademia.edu

Figure 1 – uploaded by TELKOMNIKA JOURNAL

See full PDFdownloadDownload figure
Figure 1. The schematic diagram of the procedures for the study In the investigation of the consistency, and local stability of Shapash explanation of the RF regressor prediction outcome, this study suggests the use of a random forest repressor (RFR). To explain the prediction outcome of RFR, the study employed LIME, SHAP, and ACV. Figure | highlights the study’s general method. In the evaluation of the explanation generated by these methods, the study used consistency, stability, and approximation.

Figure 1 The schematic diagram of the procedures for the study In the investigation of the consistency, and local stability of Shapash explanation of the RF regressor prediction outcome, this study suggests the use of a random forest repressor (RFR). To explain the prediction outcome of RFR, the study employed LIME, SHAP, and ACV. Figure | highlights the study’s general method. In the evaluation of the explanation generated by these methods, the study used consistency, stability, and approximation.

Related Figures (6)

Figure 2. Consistency of explainability methods The consistency metric compares how close the explanations are to each other. It evaluates the similarity of explanations generated from different explainability methods. The similarity between the explanations is determined based on the average distance between the generated explanations. Figure 2 shows the consistency among LIME, SHAP, and ACV.
Figure 2. Consistency of explainability methods The consistency metric compares how close the explanations are to each other. It evaluates the similarity of explanations generated from different explainability methods. The similarity between the explanations is determined based on the average distance between the generated explanations. Figure 2 shows the consistency among LIME, SHAP, and ACV.
Figure 3. The praise-wise comparison of consistency between tree and sampling SHAP The consistency metric demonstrated in Figure 2 shows the similarity between explanations generated by different explainability methods. The consistency is determined based on the average distance between the generated explanations by various explainability methods. The explanation between LIME, and SHAP generated a similar explanation (distance=0.35), compared to the ACV explanation (distance=0.43). In conclusion, for this particular sample, SHAP, and LIME are more similar than ACV. Moreover, Figure 3 demonstrates the consistency of the explainability methods pairwise plot for the explanation generated by different explanation methods.
Figure 3. The praise-wise comparison of consistency between tree and sampling SHAP The consistency metric demonstrated in Figure 2 shows the similarity between explanations generated by different explainability methods. The consistency is determined based on the average distance between the generated explanations by various explainability methods. The explanation between LIME, and SHAP generated a similar explanation (distance=0.35), compared to the ACV explanation (distance=0.43). In conclusion, for this particular sample, SHAP, and LIME are more similar than ACV. Moreover, Figure 3 demonstrates the consistency of the explainability methods pairwise plot for the explanation generated by different explanation methods.
Table 1. Differences in contribution distributed across HD features
Table 1. Differences in contribution distributed across HD features
Figure 4. The average distance between multiple explanations
Figure 4. The average distance between multiple explanations
Figure 5. The average distance between explanation
Figure 5. The average distance between explanation
Figure 6. The average distance between multiple explanations
Figure 6. The average distance between multiple explanations
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved