Merge pull request #202208 from ssalgadodev/AutoMLGaurdrailUpdate

Automl | gaurdrail addition plus clarity update | How to configure Auto Features

Author: PRMerger10

articles/machine-learning/how-to-configure-auto-features.md

@@ -120,11 +120,16 @@ The following table describes the data guardrails that are currently supported a
 Guardrail|Status|Condition for trigger
 ---|---|---
 **Missing feature values imputation** |Passed <br><br><br> Done| No missing feature values were detected in your training data. Learn more about [missing-value imputation.](./how-to-use-automated-ml-for-ml-models.md#customize-featurization) <br><br> Missing feature values were detected in your training data and were imputed.
-**High cardinality feature handling** |Passed <br><br><br> Done| Your inputs were analyzed, and no high-cardinality features were detected. <br><br> High-cardinality features were detected in your inputs and were handled.
+**High cardinality feature detection** |Passed <br><br><br> Done| Your inputs were analyzed, and no high-cardinality features were detected. <br><br> High-cardinality features were detected in your inputs and were handled.
 **Validation split handling** |Done| The validation configuration was set to `'auto'` and the training data contained *fewer than 20,000 rows*. <br> Each iteration of the trained model was validated by using cross-validation. Learn more about [validation data](./how-to-configure-auto-train.md#training-validation-and-test-data). <br><br> The validation configuration was set to `'auto'`, and the training data contained *more than 20,000 rows*. <br> The input data has been split into a training dataset and a validation dataset for validation of the model.
 **Class balancing detection** |Passed <br><br><br><br>Alerted <br><br><br>Done | Your inputs were analyzed, and all classes are balanced in your training data. A dataset is considered to be balanced if each class has good representation in the dataset, as measured by number and ratio of samples. <br><br> Imbalanced classes were detected in your inputs. To fix model bias, fix the balancing problem. Learn more about [imbalanced data](./concept-manage-ml-pitfalls.md#identify-models-with-imbalanced-data).<br><br> Imbalanced classes were detected in your inputs and the sweeping logic has determined to apply balancing.
 **Memory issues detection** |Passed <br><br><br><br> Done |<br> The selected values (horizon, lag, rolling window) were analyzed, and no potential out-of-memory issues were detected. Learn more about time-series [forecasting configurations](./how-to-auto-train-forecast.md#configuration-settings). <br><br><br>The selected values (horizon, lag, rolling window) were analyzed and will potentially cause your experiment to run out of memory. The lag or rolling-window configurations have been turned off.
-**Frequency detection** |Passed <br><br><br><br> Done |<br> The time series was analyzed, and all data points are aligned with the detected frequency. <br> <br> The time series was analyzed, and data points that don't align with the detected frequency were detected. These data points were removed from the dataset. 
+**Frequency detection** |Passed <br><br><br><br> Done |<br> The time series was analyzed, and all data points are aligned with the detected frequency. <br> <br> The time series was analyzed, and data points that don't align with the detected frequency were detected. These data points were removed from the dataset.
+**Cross validation** |Done| In order to accurately evaluate the model(s) trained by AutoML, we leverage a dataset that the model is not trained on. Hence, if the user doesn't provide an explicit validation dataset, a part of the training dataset is used to achieve this. For smaller datasets (fewer than 20,000 samples), cross-validation is leveraged, else a single hold-out set is split from the training data to serve as the validation dataset. Hence, for your input data we leverage cross-validation with 10 folds, if the number of training samples are fewer than 1000, and 3 folds in all other cases.
+**Train-Test data split** |Done| In order to accurately evaluate the model(s) trained by AutoML, we leverage a dataset that the model is not trained on. Hence, if the user doesn't provide an explicit validation dataset, a part of the training dataset is used to achieve this. For smaller datasets (fewer than 20,000 samples), cross-validation is leveraged, else a single hold-out set is split from the training data to serve as the validation dataset. Hence, your input data has been split into a training dataset and a holdout validation dataset.
+**Time Series ID detection** |Passed <br><br><br><br> Fixed | <br> The data set was analyzed, and no duplicate time index were detected. <br> <br> Multiple time series were found in the dataset, and the time series identifiers were automatically created for your dataset.
+**Time series aggregation** |Passed <br><br><br><br> Fixed | <br> The dataset frequency is aligned with the user specified frequency. No aggregation was performed. <br> <br> The data was aggregated to comply with user provided frequency.
+**Short series handling** |Passed <br><br><br><br> Fixed | <br> Automated ML detected enough data points for each series in the input data to continue with training. <br> <br> Automated ML detected that some series did not contain enough data points to train a model. To continue with training, these short series have been dropped or padded.
 
 ## Customize featurization