Questions tagged [shapley-value]
The shapley-value tag has no summary.
102 questions
3
votes
1
answer
22
views
Local interpretation of predictions from joint models for longitudinal and survival data
I’m currently working in a survival analysis setting with the goal of risk prediction.
In earlier work, I used static (baseline) covariates together with Random Forest–based survival models, and I ...
- 31
0
votes
0
answers
24
views
Why do Random Forest feature importance, permutation importance, and SHAP produce different feature rankings, and which one should be trusted? [closed]
I am analyzing feature importance for a supervised classification problem using a Random Forest model. I evaluated feature importance using three commonly used approaches:
Built-in Random Forest ...
2
votes
2
answers
98
views
Interpreting Shapley values for variance decomposition?
I trained a SVM multiple regression model and want to know how much each feature contributes to the prediction variance (quantified by the RMSE). I got the Shapley values for each feature on data from ...
- 209
2
votes
0
answers
40
views
Is it theoretically valid to compare SHAP values between training and test data to interpret model-driven changes in demand forecasts?
I’m currently working on a demand forecasting problem for renewed (updated) products.
The goal is to understand why the predicted demand for a renewed product differs from its previous version (before ...
- 55
2
votes
1
answer
88
views
Is SHAP value only applicable for independent features?
I want to implement SHAP for my model, but does it only applicable for set of independent features?
- 191
1
vote
1
answer
90
views
Is it correct to use SHAP to explain actual observed values?
I have a tree-based model trained for demand forecasting and SHAP is the method chosen for explaining predictions. Among the features are history lags, promotions, pricing, resizing and many demand ...
- 231
0
votes
0
answers
86
views
Interpreting Multicollinear Models with SHAP: Challenges with XGBoost and Isotonic Regression
I am familiar with SHAP and often use it when developing or assessing ML models. I want to use SHAP in a new context. I'm working on a project that relies on an XGBoost Classifier, which outputs ...
- 1
1
vote
1
answer
304
views
SHAP values across different groups
I developed and compared four ML models via Random Forest, Support Vector Machine, Logistic Regression, and Xgboost (tidymodels R package) algorithms using data without stratification by age groups. ...
2
votes
0
answers
107
views
Are SHAP values for dependent features also dependent?
If the predictors are dependent (and possibly but not necessarily correlated), are the resulting SHAP values independent?
From this answer my understanding is that for ...
- 163
1
vote
1
answer
106
views
How shap calculates the movement amount of target for a specific variable value of an individual prediction?
In shap waterfall (or force_plot) for each individual prediction, the specific variable value moves the probability or target to a certain amount, e.g. ...
- 685
1
vote
1
answer
618
views
How to interpret SHAP Values
AI newbie here. I have this MNIST image with dimensions (28, 28, 3) and a few questions in mind. It would be great if anyone could advise me on this:
How does SHAP determine the value range? For ...
3
votes
2
answers
214
views
Applying Shapley values to classification
According to the definition by Štrumbelj and Kononenko (2013), Shapley values are defined for regression predictions.
They should, however, also be applicable to classifications when the classifier ...
- 6,265
4
votes
1
answer
194
views
Kernel SHAP - Estimation of contribution function
I have been reading the online book by Christoph Molnar on Interpretable ML (link)
If we have
$$\sum_{S \subseteq M} \bigg((v(S) - (\phi_0 + \sum_{j \in S} \phi_j)\bigg)^2 K(M,S)
$$
where $S$ is set ...
- 1,240
4
votes
1
answer
403
views
Can you sum the SHAP values of multiple lagged variables?
I want to assess feature importance of a gradient boosting machine model made to forecast the count outcome $Y$. The data is a time series and I'm introducing lagged forms of the predictors in the ...
3
votes
1
answer
225
views
Why the contribution of a categorical value in SHAP trained on Catboost differs from observation to observation
Context
Let's imagine I am interested in predicting sepal length in the iris dataset using catboost.
Objective
My main objective is understanding the effect of each categorical value for ...
- 158