Forecast Override Recommendations
A Machine Learning-based application that generates data and provides recommendations that advise against potentially detrimental forecast overrides.
Related dashboards
There a no related dashboards. This functionality enables the Forecast Override Recommendations messages of the Plan Intelligence feature in Servigistics.
ETLs
Use these
ETLs to get output from Zeppelin Notebooks, transform data if required, and save results in the Servigistics schema.
Input
N/A
Output
◦ QO_ETL_PAI_ML_FOI_TempFcstDetailAndFeatureImportance_ORACLE
◦ QO_ETL_PAI_ML_FOI_TempFcstDetailAndFeatureImportance_MSSQL
◦ QO_ETL_PAI_ML_FOI_FcstDetailAndFeatureImportance_ORACLE
◦ QO_ETL_PAI_ML_FOI_FcstDetailAndFeatureImportance_MSSQL
Configurable Parameters in Zeppelin
These parameters can be configured in Zeppelin:
Parameter | Description | Default Value |
|---|
train_split_size | The ratio for splitting the base data into training and test sets. For example, a value of 0.75 allocates 75% of the available data for model training, leaving the remaining 25% for testing. The performance on the test data is used to determine the most optimal model. | 0.75 |
feature_threshold | This is used for feature selection. The cumulative sum of feature importance, utilized during model training, will not exceed this threshold. | 0.8 |
col_missing_value_threshold | Features with a percentage of missing values surpassing this threshold are eliminated. | 0.5 |
override_recomm_threshold | Recommendations are generated for SKUs with confidence levels exceeding this threshold. | 0.8 |