Read ARFF (Advanced File Connectors)

Synopsis

This operator is used for reading an ARFF file.

Description

This operator can read ARFF (Attribute-Relation File Format) files known from the machine learning library Weka. An ARFF file is an ASCII text file that describes a list of instances sharing a set of attributes. ARFF files were developed by the Machine Learning Project at the Department of Computer Science of The University of Waikato for use with the Weka machine learning software. Please study the attached Example Process for understanding the basics and structure of the ARFF file format. Please note that when an ARFF file is written, the roles of the attributes are not stored. Similarly when an ARFF file is read, the roles of all the attributes are set to regular.

Input

file
An ARFF file is expected as a file object which can be created with other operators with file output ports like the Read File operator.

Output

output (Data table)
This port delivers the ARFF file in tabular form along with the meta data. This output is similar to the output of the Retrieve operator.

Parameters

data fileThe path of the ARFF file is specified here. It can be selected using the choose a file button.
encodingThis is an expert parameter. A long list of encoding is provided; users can select any of them.
read not matching values as missingsThis is an expert parameter. If this parameter is set to true, values that do not match with the expected value type are considered as missing values and are replaced by '?'. For example if 'abc' is written in an integer column, it will be treated as a missing value. Question mark (?) in ARFF file is also read as missing value.
decimal characterThis character is used as the decimal character.
grouped digitsThis parameter decides whether grouped digits should be parsed or not. If this parameter is set to true, the grouping character parameter should be specified.
grouping characterThis parameter is available only when the grouped digits parameter is set to true.This character is used as the grouping character. If it is found between numbers, the numbers are combined and this character is ignored. For example if "22-14" is present in the ARFF file and "-" is set as grouping character, then "2214" will be stored.
infinity representationThis parameter can be set to parse a specific infinity representation (e.g. "Infinity"). If it is not set, the local specific infinity representation will be used.

Tutorial Processes

The basics of the ARFF

The 'Iris' data set is loaded using the Retrieve operator. The Write ARFF operator is applied on it to write the 'Iris' data set into an ARFF file. The example set file parameter is set to 'D:\Iris'. Thus an ARFF file is created in the 'D' drive of your computer with the name 'Iris'. Open this file to see the structure of an ARFF file.

ARFF files have two distinct sections. The first section is the Header information, which is followed by the Data information. The Header of the ARFF file contains the name of the Relation and a list of the attributes. The name of the Relation is specified after the @RELATION statement. The Relation is ignored. Each attribute definition starts with the @ATTRIBUTE statement followed by the attribute name and its type. The resultant ARFF file of this Example Process starts with the Header. The name of the relation is 'RapidMinerData'. After the name of the Relation, six attributes are defined.

Attribute declarations take the form of an ordered sequence of @ATTRIBUTE statements. Each attribute in the data set has its own @ATTRIBUTE statement which uniquely defines the name of that attribute and its data type. The order of declaration of the attributes indicates the column position in the data section of the file. For example, in the resultant ARFF file of this Example Process the 'label' attribute is declared at the end of all other attribute declarations. Therefore values of the 'label' attribute are in the last column of the Data section.

The possible attribute types in ARFF are:

numeric
integer
real
{nominalValue1,nominalValue2,...} for nominal attributes
string for nominal attributes without distinct nominal values (it is however recommended to use the nominal definition above as often as possible)
date [date-format] (currently not supported)

You can see in the resultant ARFF file of this Example Process that the attributes 'a1', 'a2', 'a3' and 'a4' are of real type. The attributes 'id' and 'label' are of nominal type. The distinct nominal values are also specified with these nominal attributes.

The ARFF Data section of the file contains the data declaration line @DATA followed by the actual example data lines. Each example is represented on a single line, with carriage returns denoting the end of the example. Attribute values for each example are delimited by commas. They must appear in the order that they were declared in the Header section (i.e. the data corresponding to the n-th @ATTRIBUTE declaration is always the n-th field of the example line). Missing values are represented by a single question mark (?).

A percent sign (%) introduces a comment and will be ignored during reading. Attribute names or example values containing spaces must be quoted with single quotes ('). Please note that the sparse ARFF format is currently only supported for numerical attributes. Please use one of the other options for sparse data files provided by RapidMiner if you also need sparse data files for nominal attributes.

Reading an ARFF file using the Read ARFF operator

The ARFF file that was written in the first Example Process using the Write ARFF operator is retrieved in this Example Process using the Read ARFF operator. The data file parameter is set to '%{tempdir}/Iris'. All other parameters are used with default values. Run the process. You will see that the results are very similar to the original Iris data set in the repository. Please note that the role of all the attributes is regular in the results of the Read ARFF operator. Even the roles of 'id' and 'label' attributes are set to regular. This is so because the ARFF files do not store information about the roles of the attributes.

Categories

Versions