Free online ARFF to CSV converter

A spreadsheet, database importer, or data-analysis program may require a flat CSV table instead of a Weka dataset. Converting it improves compatibility and editing, but CSV cannot retain all schema information stored in the ARFF header.

What the ARFF format is

ARFF (Attribute-Relation File Format) is a plain-text dataset format used primarily by Weka and Weka-compatible machine-learning tools. A typical .arff file contains a @RELATION declaration, @ATTRIBUTE definitions, and a @DATA section containing records.

Attribute declarations can define numeric, nominal, string, date, and Weka relational values. For example, @ATTRIBUTE class {yes,no} identifies a categorical field and lists its permitted values. ARFF also supports sparse records, comments, missing values commonly represented by ?, and other metadata that describes the dataset.

Weka users encounter ARFF when downloading machine-learning datasets, exporting data from Weka Explorer, or exchanging datasets with software that understands the format.

What the CSV format is

CSV (comma-separated values) represents a table as records on separate lines, with fields separated by a delimiter, normally a comma. The first row often contains column names, although CSV does not require a header or define a universal schema.

Spreadsheet applications, database importers, Python, R, statistical software, and data portals commonly accept CSV. It is easy to inspect and edit as text, but it has no standard place for ARFF relation names, declared data types, nominal-value lists, class attributes, comments, or sparse-row notation.

Convert with Weka

1. Install and open Weka, then start Explorer.
2. On the Preprocess tab, select Open file... and choose the .arff file.
3. Check the displayed attributes, types, instance count, class setting, and missing values.
4. Select Save..., choose the CSV data-file format in the save dialog if necessary, and save the result as output.csv.
5. Open the CSV in a text editor or its destination program and verify the header, row count, delimiter, encoding, and missing-value representation.

For repeatable conversions, run Weka's CSV saver from a terminal. Replace the JAR path with the location of the Weka installation:

java -cp /path/to/weka.jar weka.core.converters.CSVSaver -i input.arff -o output.csv

Quote Windows paths containing spaces, for example "C:\Program Files\Weka-3-8\weka.jar". A Weka launcher or installation-specific classpath may be required with some versions.

Convert with Python

pandas does not natively parse ARFF. For a standard dense file, install an ARFF parser and pandas, then create a DataFrame from the parsed attributes and records:

pip install liac-arff pandas

import arff
import pandas as pd

with open("input.arff", encoding="utf-8") as source:
    dataset = arff.load(source)

columns = [name for name, _ in dataset["attributes"]]
pd.DataFrame(dataset["data"], columns=columns).to_csv("output.csv", index=False)

Test the result carefully when the file contains sparse instances, relational attributes, unusual quoting, dates, or parser-specific features. Weka is generally the safer choice for Weka-specific ARFF files.

Use the built-in online converter

You can convert a file on 101convert.com for free without installing software: upload the ARFF file, select CSV as the output format, start the conversion, and download the result. Do not upload confidential, personal, medical, financial, or proprietary datasets unless the service's current handling and deletion terms meet your requirements.

Check the converted file

• Schema is not preserved: keep the original ARFF file or record the relation name, attribute types, nominal-value lists, class attribute, date formats, and other metadata separately.
• Missing values may change meaning: ARFF commonly uses ?, while a destination may expect an empty field, NA, or NULL. Confirm the destination's import settings before replacing any marker.
• Sparse rows require expansion: a CSV row must contain a field for every column, so a converter expands sparse ARFF records according to ARFF parsing rules. Verify omitted values, especially in nonnumeric attributes.
• Values are quoted according to CSV rules: fields containing commas, quotation marks, or line breaks must be escaped and quoted. Do not convert by replacing commas or splitting lines with basic text commands.
• Types must be reassigned: numeric, date, and nominal values become untyped CSV fields. A receiving application may infer them incorrectly, and a nominal value such as yes no longer carries its declared category list.
• Encoding and delimiters vary: retain UTF-8 for non-ASCII text, and use the destination application's import dialog to select comma, semicolon, or another delimiter if the whole file appears in one column.