Appendix
Appendix A. Input and Output Data Sets¶
There are two input data sets: one for the Training phase and one for the Execution phase.
There are seven output data sets: four for the Training phase and three for the Execution phase.
It is recommended to set each numerical field in the following output data sets to five decimal places to avoid errors caused by large numerical values.
Starting with Tagetik version SP31, it is possible to use data sources as input for both the Training and Execution phases.
A.1 Input Training Data Set¶
This data set contains historical data:
- OID: This is the first field, and it identifies the row ID (it does not need to be mapped; it is just present in the data)
- REF_DATE_1: The first field that identifies the time of the series (text)
- REF_DATE_2: The second field that identifies the time of the series (text)
- …
- AMOUNT: The amount for each series (number)
- DIMENSIONS_0: The first field that identifies the time series
- DIMENSIONS_1: The second field that identifies the time series
- ….
- MEASURE: The name of the variable
- TARGET_WEIGHTS: The weight (from 0 to 1) used to identify target variables and to associate a weight during validation for the datapoint
The REF_DATE and DIMENSION columns may have a variable number of columns depending on the context.
For example:
| OID | DIMENSIONS_0 | TARGET_WEIGHTS | REF_DATE_1 | REF_DATE_2 | AMOUNT | MEASURE |
|---|---|---|---|---|---|---|
| ab23c617 | EN_01 | 1 | 2025_ACTUAL | 01 | 7530 | VOLUME |
| ab23c618 | EN_01 | 1 | 2025_ACTUAL | 02 | 8593 | VOLUME |
| ab23c640 | EN_01 | 1 | 2025_ACTUAL | 03 | 9850 | VOLUME |
| … | … | … | … | … | … | … |
A.2 Input Execution Data set¶
This data set contains data to use for prediction:
- OID: This is the first field, and it identifies the row ID (it does not need to be mapped; it is just present in the data)
- REF_DATE_1: The first field that identifies the time of the series (text)
- REF_DATE_2: The second field that identifies the time of the series (text)
- …
- AMOUNT: The amount for each series (number)
- DIMENSIONS_0: The first field that identifies the time series
- DIMENSIONS_1: The second field that identifies the time series
- ….
- MEASURE: The name of the variable
The REF_DATE and DIMENSION columns must be defined according to the training step.
For example:
| OID | DIMENSIONS_0 | TARGET_WEIGHTS | REF_DATE_1 | REF_DATE_2 | AMOUNT | MEASURE |
|---|---|---|---|---|---|---|
| ab23c617 | EN_01 | 1 | 2026_ACTUAL | 01 | 8510 | VOLUME |
| ab23c618 | EN_01 | 1 | 2026_ACTUAL | 02 | 9892 | VOLUME |
| ab23c640 | EN_01 | 1 | 2026_ACTUAL | 03 | 9979 | VOLUME |
| … | … | … | … | … | … | … |
A.3 Output Training Data Set for Metrics¶
This data set contains the metrics of the model:
- OID: This is the first field, and it identifies the row ID (it does not need to be mapped; it is just present in the data)
- AMOUNT: The metric value for each series (number)
- DIMENSIONS_0: The first field that identifies the time series
- DIMENSIONS_1: The second field that identifies the time series
- ….
- MEASURE: The name of the variable
- METRIC: The metric (text)
For example:
| OID | DIMENSIONS_0 | AMOUNT | METRIC | MEASURE |
|---|---|---|---|---|
| ab23c617 | EN_01 | 0.045 | MAPE | VOLUME |
| ab23c700 | EN_01 | 0.140 | R2 | SALES |
| … | … | … | … | … |
A.4 Output Training Data Set for Fit
This data set contains the fit of the model:
- OID: This is the first field, and it identifies the row ID (it does not need to be mapped; it is just present in the data)
- REF_DATE_1: The first field that identifies the time of the series (text)
- REF_DATE_2: The second field that identifies the time of the series (text)
- …
- FIT: The fit amount for each series (number)
- DIMENSIONS_0: The first field that identifies the time series
- DIMENSIONS_1: The second field that identifies the time series
- ….
- MEASURE: The name of the variable
For example:
| OID | DIMENSIONS_0 | REF_DATE_1 | REF_DATE_2 | FIT | MEASURE |
|---|---|---|---|---|---|
| ab23c617 | EN_01 | 2025_ACTUAL | 01 | 8610 | VOLUME |
| ab23c618 | EN_01 | 2025_ACTUAL | 02 | 9800 | VOLUME |
| ab23c640 | EN_01 | 2025_ACTUAL | 03 | 9974 | VOLUME |
| … | … | … | … | … | … |
A.5 Output Training Data Set for Importance¶
This data set contains the forecast of the model:
- OID: This is the first field, and it identifies the row ID (it does not need to be mapped; it is just present in the data)
- IMPORTANCE: The importance value of an external feature for each series (number)
- DIMENSIONS_0: The first field that identifies the time series
- DIMENSIONS_1: The second field that identifies the time series
- ….
- MEASURE_TARGET: The name of the target variable
- VARIABLE: The name of the external feature (text)
- LAG: The lag of the external feature (text)**
** If the variable (external feature) is the seasonality feature, the lag column contains the seasonality information for that variable. If the variable is a dimension feature, the lag column remains blank.
For example:
| OID | DIMENSIONS_0 | VARIABLE | LAG | IMPORTANCE | MEASURE_TARGET |
|---|---|---|---|---|---|
| ab23c617 | EN_01 | PRICE | -1 | 0.30 | VOLUME |
| ab23c618 | EN_01 | PRICE_SEASONALITY | 6 | 0.25 | VOLUME |
| ab23c640 | EN_01 | DIMENSIONS_0 | 0.45 | VOLUME |
A.6 Output Training Data Set for Contribution¶
This data set contains the contributions of the exogenous variable of the model:
- OID: This is the first field, and it identifies the row ID (it does not need to be mapped; it is just present in the data)
- REF_DATE_1: The first field that identifies the time of the series (text)
- REF_DATE_2: The second field that identifies the time of the series (text)
- …
- CONTRIBUTION: The contribution value of an external feature for each series (number)
- DIMENSIONS_0: The first field that identifies the time series
- DIMENSIONS_1: The second field that identifies the time series
- ….
- MEASURE_TARGET: The name of the target variable
- VARIABLE: The name of the external feature (text)
- LAG: The lag of the external feature (text)**
** If the variable (external feature) is the seasonality feature, the lag column contains the seasonality information for that variable. If the variable is a dimension feature, the lag column remains blank.
For example:
| OID | DIMENSIONS_0 | VARIABLE | LAG | CONTRIBUTION | REF_DATE_0 | REF_DATE_1 | MEASURE_TARGET |
|---|---|---|---|---|---|---|---|
| ab23c617 | EN_01 | PRICE | -1 | 200 | 2025_ACTUAL | 01 | VOLUME |
| ab23c618 | EN_01 | PRICE_SEASONALITY | 6 | 60 | 2025_ACTUAL | 01 | VOLUME |
| ab23c640 | EN_01 | DIMENSIONS_0 | 15 | 2025_ACTUAL | 01 | VOLUME | |
| … | … | … | … | … | … |
A.7 Output Execution Data Set for Forecast¶
This data set contains the forecast of the model:
- OID: This is the first field, and it identifies the row ID (it does not need to be mapped; it is just present in the data)
- REF_DATE_1: The first field that identifies the time of the series (text)
- REF_DATE_2: The second field that identifies the time of the series (text)
- …
- FORECAST: The forecasted amount for each series (number)
- DIMENSIONS_0: The first field that identifies the time series
- DIMENSIONS_1: The second field that identifies the time series
- ….
- MEASURE: The name of the variable
- LOWER: The lower bound of the confidence interval
- UPPER: The upper bound of the confidence interval
- IS_OUTLIER_PREDICTION: The prediction of outlier/inlier
- OUTLIER_PROBABILITY: The probability of being an outlier
For example:
| OID | DIMENSIONS_0 | REF_DATE_1 | REF_DATE_2 | FORECAST | MEASURE | UPPER | LOWER | IS OUTLIER PREDICTION | OUTLIER PROBABILITY |
|---|---|---|---|---|---|---|---|---|---|
| ab23c617 | EN_01 | 2025_ACTUAL | 01 | 8610 | VOLUME | 1000 | 7220 | 1 | 0.7 |
| ab23c618 | EN_01 | 2025_ACTUAL | 02 | 9800 | VOLUME | 1200 | 7600 | 0 | 0.1 |
| ab23c640 | EN_01 | 2025_ACTUAL | 03 | 9974 | VOLUME | 1300 | 6.948 | 0 | 0.2 |
| … | … | … | … | … | .. | … | … |
A.8 Output Execution Data Set for Importance¶
This data set contains the forecast of the model:
- OID: This is the first field, and it identifies the row ID (it does not need to be mapped; it is just present in the data)
- IMPORTANCE: The importance value of an external feature for each series (number)
- DIMENSIONS_0: The first field that identifies the time series
- DIMENSIONS_1: The second field that identifies the time series
- ….
- MEASURE_TARGET: The name of the target variable
- VARIABLE: The name of the external feature (text)
- LAG: The lag of the external feature (text)**
** If the variable (external feature) is the seasonality feature, the lag column contains the seasonality information for that variable. If the variable is a dimension feature, the lag column remains blank.
For example:
| OID | DIMENSIONS_0 | VARIABLE | LAG | IMPORTANCE | MEASURE_TARGET |
|---|---|---|---|---|---|
| ab23c617 | EN_01 | PRICE | -1 | 0.50 | VOLUME |
| ab23c618 | EN_01 | PRICE_SEASONALITY | 6 | 0.15 | VOLUME |
| ab23c640 | EN_01 | DIMENSIONS_0 | 0.35 | VOLUME | |
| … | … | … | … | … | … |
A.9 Output Execution Data Set for Contribution¶
This data set contains the contribution of each variable and date in the prediction.
- OID: This is the first field, and it identifies the row ID (it does not need to be mapped; it is just present in the data)
- REF_DATE_1: The first field that identifies the time of the series (text)
- REF_DATE_2: The second field that identifies the time of the series (text)
- …
- CONTRIBUTION: The contribution value of an external feature for each series (number)
- DIMENSIONS_0: The first field that identifies the time series
- DIMENSIONS_1: The second field that identifies the time series
- ….
- MEASURE_TARGET: The name of the target variable
- VARIABLE: The name of the external feature (text)
- LAG: The lag of the external feature (text)**
** If the variable (external feature) is the seasonality feature, the lag column contains the seasonality information for that variable. If the variable is a dimension feature, the lag column remains blank.
For example:
| OID | REF_DATE_0 | REF_DATE_1 | DIMENSIONS_0 | VARIABLE | LAG | CONTRIBUTION | MEASURE_TARGET |
|---|---|---|---|---|---|---|---|
| ab23c617 | 2025_ACTUAL | 01 | EN_01 | PRICE | -1 | 100 | VOLUME |
| ab23c618 | 2025_ACTUAL | 01 | EN_01 | PRICE_SEASONALITY | 6 | 50 | VOLUME |
| ab23c640 | 2025_ACTUAL | 01 | EN_01 | DIMENSIONS_0 | 25 | VOLUME | |
| … | … | … | … | … | … |
Appendix B. Dimensions and Reference Date¶
DIMENSIONS LIST: If you need multiple dimensions to identify the time series key (e.g., ENTITY, CHANNEL, PRODUCT), you can map the fields directly in the configuration window.
REF_DATE: If you need multiple dimensions to identify the time ID for the historical series (e.g., SCENARIO, PERIOD), you can map the fields directly in the configuration window.
For example:
Appendix C. Parameters¶
The parameters to set for the Training phase include the following:
- LOG LEVEL:
- DEBUG: all information logged
- INFO: all warnings, errors, critical errors, and general information logged without debugging information
- WARNING: all warnings, errors, and critical errors are logged
- ERROR: all errors and critical errors are logged
- CRITICAL: only critical errors are logged
Logging messages that are less severe than the specified level are ignored, while messages with a severity level or higher are logged.
- FEATURES_MAX_LAG: Maximum lag applicable
- USE_TARGET_FEATURE_SEASONALITY: whether to create the seasonal feature of the target or not.
- USE_EXOGENOUS_FEATURES_SEASONALITY: whether to create or omit the seasonality features of the exogenous variables.
- LAG_TARGET_FEATURES: whether to generate lagged features of the target features.
- LAG_EXOGENOUS_FEATURES: whether to create lagged features of exogenous variables or not.
- GET_CONTRIBUTIONS: used to retrieve driver contributions as output.
- GET_IMPORTANCES: to obtain the driver’s importance as output.
- SCALE_FEATURES: to scale the input features before training the neural network.
- FEATURES_SELECTION_THRESHOLD: (from 0 to 1) the threshold used to determine whether a feature should be excluded based on its importance.
- INCLUDE_OTHERS_TARGETS_IMPACT: to include the lags of other targets as features for a target with the same dimensionality.
- Use the lagged exogenous features: Y (Yes) or N (No) to create the lagged version of the exogenous features. The default is Y (Yes), but you can set it to N (No) to reduce processing time.
- Feature max lag: the maximum lag value that can be applied to the selected variables (target or exogenous). The default is AUTO, which corresponds to the seasonal value.
- Use the lagged target feature: Y (Yes) or N (No) to create the lagged version of the target features. The default is Y (Yes), but you can set it to N (No) to reduce processing time.
The parameters to set for the Execution phase include:
- LOG LEVEL:
- DEBUG: all information logged
- INFO: all warnings, errors, critical errors, and general information logged without debugging information
- WARNING: all warnings, errors, and critical errors are logged
- ERROR: all errors and critical errors are logged
- CRITICAL: only critical errors are logged
Logging messages that are less severe than the specified level are ignored, while messages with a severity level or higher are logged.
- CONFIDENCE_LEVEL: the confidence level used to calculate the confidence interval.
- GET_CONTRIBUTIONS: used to retrieve driver contributions as output.
- GET_IMPORTANCES: to obtain the driver’s importance as output.
- EXECUTE_ANOMALY: to perform anomaly detection on the target having actual values as input execute.