This node helps handle missing values found in cells of the input table.
The first section "Treatment by Column" permits individual settings for each available column. Only columns that are not listed here are treated according to the settings in the second section "Treatment by Data Type". It is possible to select more than one column within one element leaving only treatment options that are applicable to all selected columns enabled.
\n \nThe second section "Treatment by Data Type" provides default handling options for all columns of a given type. These settings apply to all columns in the input table that are not explicitly mentioned in the first section "Treatment by Column".
\n \nTreatment options marked with an asterisk (*) will result in non-standard PMML, which uses extensions that cannot be read by other tools than KNIME.
\n \n \n ","shortDescription":"Filters or replaces missing values in a table.","keywords":["Imputation"],"deprecated":false,"streamable":false,"tags":["Manipulation","Column","Transform"],"icon":"data:image/png;base64,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","views":[],"inPorts":[{"objectClass":"org.knime.core.node.BufferedDataTable","color":"000000","dataType":"Table","optional":false,"name":"Input table","description":"Table with missing values."}],"outPorts":[{"objectClass":"org.knime.core.node.BufferedDataTable","color":"000000","dataType":"Table","optional":false,"name":"Output table","description":"Table with replaced missing values."},{"objectClass":"org.knime.core.node.port.pmml.PMMLPortObject","color":"1469af","dataType":"PMML","optional":false,"name":"PMML Transformations","description":"Table with PMML documenting the missing value replacement."}],"dynInPorts":[],"dynOutPorts":[],"additionalSiteInformation":{"url":"https://update.knime.com/analytics-platform/5.12","trusted":true,"enabledByDefault":true,"name":"KNIME Analytics Platform 5.12 Update Site","version":"5.12.0"},"dialog":[{"fields":[{"name":"Missing Value Handler Selection","description":"Select and configure the missing value handler to be used for data types or columns. Handlers that do not produce valid PMML 4.2 are marked with an asterisk (*).\nAverage Interpolation*\n
\n This missing value handler replaces missing values with the average value of the previous and next encountered non-missing value in the column it is configured for. When dealing with tables that have a large number of rows but not too many columns that need missing value replacement, the option to use disk backed statistics avoids flooding of the main memory. This should be used with caution, as it is generally much slower than in-memory statistics. This missing value handler does not produce standard PMML 4.2!
Fix Value (Double)\n
\n Replaces missing values with a double given by the user. This missing value handler produces valid PMML 4.2.
Fix Value (Integer)\n
\n Replaces missing values with an integer number given by the user. This missing value handler produces valid PMML 4.2.
Fix Value (String)\n
\n Replaces missing values with a string given by the user. This missing value handler produces valid PMML 4.2.
Fix Value (Long)\n
\n Replaces missing values with a long given by the user. This missing value handler produces valid PMML 4.2.
Fix Value\n
\n No description provided.
Linear Interpolation*\n
\n This missing value handler replaces missing values with the linear interpolation between the previous and next encountered non-missing value in the column it is configured for. When dealing with tables that have a large number of rows but not too many columns that need missing value replacement, the option to use disk backed statistics avoids flooding of the main memory. This should be used with caution, as it is generally much slower than in-memory statistics. This missing value handler does not produce standard PMML 4.2!
Maximum\n
\n Finds the column's largest value and replaces all missing values with it. This missing value handler produces valid PMML 4.2.
Mean\n
\n Calculates the mean value of all non-missing cells in a column and replaces the missing values with this mean. This missing value handler produces valid PMML 4.2.
Median\n
\n Finds the column's median value and replaces all missing values with it. For large tables this might be computationally expensive because the table needs to be sorted to find the median. This missing value handler produces valid PMML 4.2.
Minimum\n
\n Finds the column's smallest value and replaces all missing values with it. This missing value handler produces valid PMML 4.2.
Most Frequent Value\n
\n Calculates the most frequent value in a column and replaces the missing values with it. This missing value handler produces valid PMML 4.2.
Moving Average*\n
\n Calculates the mean of all values that are within the window given by the lookahead and lookbehind and replaces missing values with this mean. This missing value handler does not produce standard PMML 4.2! The number of cells to take into account before and after the current cell can be set using the options lookbehind and lookahead respectively.
Next*\n
\n This missing value handler replaces missing values with the next encountered non-missing value in the column it is configured for. When dealing with tables that have a large number of rows but not too many columns that need missing value replacement, the option to use disk backed statistics avoids flooding of the main memory. This should be used with caution, as it is generally much slower than in-memory statistics. This missing value handler does not produce standard PMML 4.2!
Previous*\n
\n This missing value handler replaces missing values with the last encountered non-missing value in the column it is configured for. This missing value handler does not produce standard PMML 4.2!
Remove Row*\n
\n This missing value handler removes rows that have a missing value in the column it is configured for. This missing value handler does not produce standard PMML 4.2!
Rounded Mean\n
\n Calculates the mean value of all non-missing cells in a column and replaces the missing values with this mean. This missing value handler produces valid PMML 4.2.