Linear regression coefficient and intercept

andres_kukke · ‎2024-09-13

Hei!

I buliding model to predict market price. I made experiment and best model was linear regression with 4 features. I trained same model in R-Studio using also linear regression (function lm())
Qlik AI gives highest R-square (0.65 vs 0.59 in R-Studio) into same sample (also same number of features)
I was curious to find out regression coefficient and intercept. In R-Studio I can get coefficients using syntax: summary(my_regression_model).

Residuals:
     Min       1Q   Median       3Q      Max 
-10.4651  -1.5801  -0.3867   1.4452   9.1223 

Coefficients:
                          Estimate Std. Error t value Pr(>|t|)    
(Intercept)                -6.6118     2.1749  -3.040  0.00367 ** 
Volume                      6.4781     1.2279   5.276 2.50e-06 ***
SalesABC                    0.8449     2.2853   0.370  0.71308    
Corrosion                   3.2238     1.1819   2.728  0.00863 ** 
Quality                     4.4110     0.9713   4.541 3.26e-05 ***
Volume:SalesABC            -1.7214     1.1693  -1.472  0.14690    
Volume:Corrosion           -1.4987     0.8246  -1.817  0.07480 .  
SalesABC:Corrosion         -0.9049     1.0123  -0.894  0.37543    
Volume:SalesABC:Corrosion   0.4691     0.6139   0.764  0.44814    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 3.225 on 53 degrees of freedom
Multiple R-squared:  0.645,	Adjusted R-squared:  0.5914 
F-statistic: 12.04 on 8 and 53 DF,  p-value: 1.323e-09

How can I get regression model coefficients and intercept in Qlik AI?
I can get out SHAP values qvd (or csv file) but those are different than regression coefficients.

Does Qlik AI supports "explainable AI" concept?

Andres

Kyle_Jourdan · ‎2024-09-17

Qlik AutoML does not expose the coefficients as they are only available for linear/logistic regression models.

In reality, most use cases are being best solved with tree-based or gradient-boosted algorithms, for which coefficients are not available. In addition, most users of AutoML do not require these coefficients for their use cases.

For consistency across all algorithm types, AutoML uses SHAP values for explainability.

Model Explanation