Please enable JavaScript.
Coggle requires JavaScript to display documents.
Ch.19 (the directional impact of a feature (whether the presence of a…
Ch.19
- the directional impact of a feature
whether the presence of a value helps the model by assisting it in predicting readmissions or non-readmissions
what will shown is the result of a logistic regression analysis. Do note, however, that the visual was developed using only a 16% sample size
The reason this analysis is only available at 16% of the data is because the model did not work well enough to be among the models selected to be retrained using 32% or 64%.
While the DataRobot Variable Effect screen does not provide all the information to recreate a model, it does provide what are commonly known as coefficients for the most important feature characteristics that drive a prediction decision.
- the partial impact of a feature
-
This will take some time to process as it must consider eight different models while calculating the values of the features
therein.
The Model X-Ray constructs a list of features ranked by
their influence on the target as denoted by the size of the green line under each feature name in the left pane. This information is the same as that used to rank features under the Data tab, but here, it displays 100% of the whole bar.
The left Y-axis contains the frequency of cases in the validation set. The X-axis contains the values of the most
predictive feature, discharge_disposition_id.
the power of language
-
-
-
for any text-heavy analysis, the option Filter Stop Words , set to yes by default , is an important detail to note.
- the overall impact of a feature adjusted for the impact of other features
click on the top model as sorted by cross validation, ENET Blender and then select Feature Impact, followed by Compute Feature Impact. ->this will initiate a calculation of the value of each feature in the context of this model.
DataRobot then examines the model’s performance relative to the model that retained all the features. The extent to which the model with the randomly shuffled feature does worse than the original model is assigned as the feature’s value. This procedure is done with every feature .
DataRobot has conveniently placed an option to select a set of top features as a new feature list. All that is necessary is to give the new list a name and select how many of the top features will populate it. A new model run can then be done using this feature list. Creating models with fewer features is generally a good idea to avoid overfitting, and can also reduce problems due to changes in the databases and sources of data.
- the overall impact of a feature without consideration of the impact of other features
The overall impact of a feature without consideration for the impact of other features treats each feature as a standalone effect on the target.
-
Reason codes
The focus of this chapter has thus far been on the features and feature values that drive positive or negative changes in the target; however, understanding of the data at the more granular individual patient level remains limited.
To address the individual data points, each patient’s feature values need to be examined along with an analysis of how these values determined that patient’s probability of readmission.
The left and right thresholds allow for the specification of the probability cutoffs to be used in this view.
-
Be aware that computing reason codes is slower than computing
predictions because reason codes actually engage in additional evaluations of why a prediction was set as the given probability for that case.
This is not likely to be a problem since reason codes are primarily of use in settings where human beings are involved in the decision process.
four kinds of relationship that are commonly useful for exploring why a model predicts certain outcomes.