A Machine Learning Model Created in Oracle Analytics Fails With 'KeyError: "['Column_Name'] not in index" (Doc ID 2705900.1)

Last updated on NOVEMBER 05, 2023

Applies to:

Symptoms

You are creating a machine learning training model to predict shortages with Oracle Analytics Server (OAS)
Note: This can apply to all Oracle Analytics family of products, such as Oracle Analytics Cloud (OAC) and Oracle Analytics Desktop (OAD)

When you apply your training model to your data set, it errors with:

Data Flow "[dataflow name]" has encountered an error.

Something Went Wrong

The error details look like this:

Step |j| Execution failed. Status: FAILED. Message: Error Message: "['COLNAME1' COLNAME2' 'COLNME3'] not in index"
Traceback (most recent call last):
File "[ORACLE_HOME]/bi/servers/obis1/tmp/nQS_PAF_31991_99_32706155.TMP", line 84, in <module>
return_dat = obi_apply_model(dat, model, columnMetadata, args)
File "[ORACLE_HOME]/bi//servers/obis1/tmp/nQS_PAF_31991_99_32706155.TMP", line 46, in obi_apply_model
return_dat = obipy.tools.linear_regression.apply_model(model, dat, args)
File "[ORACLE_HOME]/bi/bifoundation/advanced_analytics/python_scripts/obipy/tools/linear_regression.py", line 39, in apply_model
df = RegularizedLinearRegression().apply_model(model, df, args, includeInput=args['includeInputColumns'])
File "[ORACLE_HOME]/bi/bifoundation/advanced_analytics/python_scripts/obipy/obiml/RegularizedLinearRegression.py", line 211, in apply_model
data = pd.concat([categorical_data, df[input_features['numerical']]], axis=1)
File "build/bdist.linux-x86_64/egg/pandas/core/frame.py", line 2133, in __getitem__
File "build/bdist.linux-x86_64/egg/pandas/core/frame.py", line 2177, in _getitem_array
File "build/bdist.linux-x86_64/egg/pandas/core/indexing.py", line 1269, in _convert_to_indexer
KeyError: "['COLNAME1' COLNAME2' 'COLNME3'] not in index"
[nQSError: 43224] The Dataflow "Untitled" failed during the execution.
[nQSError: 43204] Asynchronous Job Manager failed to execute the asynchronous job.

Cause

My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.