Description:

A thin Python wrapper that uses the javabridge Python library to communicate with a Java Virtual Machine executing Weka API calls. Offers all major APIs, like data generators, loaders, savers, filters, classifiers, clusterers, attribute selection, associations and experiments. Weka packages can be listed/installed/uninstalled as well. It does not provide any graphical frontend, but some basic plotting and graph visualizations are available through matplotlib and pygraphviz. A simple workflow engine was added with release 0.3.0.

Changes to previous version:

• upgraded to Weka 3.9.2
• properly initializing package support now, rather than adding package jars to classpath
• added weka.core.ClassHelper Java class for obtaining classes and static fields, as javabridge only uses the system class loader

BibTeX Entry: Download

Supported Operating Systems: Agnostic

Data Formats: Arff, Csv, Libsvm, Xrff

Tags: Machine Learning, Weka

Archive: download here

Other available revisons

Version	Changelog	Date
0.3.12	upgraded to Weka 3.9.2 properly initializing package support now, rather than adding package jars to classpath added weka.core.ClassHelper Java class for obtaining classes and static fields, as javabridge only uses the system class loader	February 18, 2018, 04:29:24
0.3.11	added check_for_modified_class_attribute method to FilterClassifier class added complete_classname method to weka.core.classes module, which allows completion of partial classnames like .J48 to weka.classifiers.trees.J48; if there is a unique match; JavaObject.new_instance and JavaObject.check_type now make use of this functionality, allowing for instantiations like Classifier(cls=".J48") jvm.start(system_cp=True) no longer fails with a KeyError: 'CLASSPATH' if there is no CLASSPATH environment variable defined Libraries mtl.jar, core.jar and arpack_combined_all.jar were added as is to the weka.jar in the 3.9.1 release instead of adding their content to it. Repackaged weka.jar to fix this issue.	August 23, 2017, 01:17:24
0.3.10	"types.double_matrix_to_ndarray" no longer assumes a square matrix (https://github.com/fracpete/python-weka-wrapper/issues/48) "len(Instances)" now returns the number of rows in the dataset (module "weka.core.dataset") added method "insert_attribute" to the "Instances" class added class method "create_relational" to the "Attribute" class upgraded Weka to 3.9.1	January 4, 2017, 10:21:33
0.3.9	plot_learning_curve method of module weka.plot.classifiers now accepts a list of test sets; * is index of test set in label template string added missing_value() methods to weka.core.dataset module and Instance class output variable y for convenience method create_instances_from_lists in module weka.core.dataset is now optional added convenience method create_instances_from_matrices to weka.core.dataset module to easily create an Instances object from numpy matrices (x and y)	October 18, 2016, 22:55:00
0.3.8	works now with javabridge 1.0.14	May 9, 2016, 04:17:42
0.3.7	upgraded Weka to 3.9.0	May 4, 2016, 00:02:47
0.3.6	Loader.load_file method now checks whether the dataset file really exists, otherwise a previously loaded file gets loaded again without an error message (seems to be a Weka issue) replaced org.pentaho.packageManagement with weka.core.packageManagement as the package management code is now part of Weka rather than a third-party library jvm.start() no longer tries to load packages and therefore suppresses error message if $HOME/wekafiles/packages should not yet exist	April 2, 2016, 11:57:09
0.3.5	added support for weka.core.BatchPredictor to class Classifier in module weka.classifiers upgraded Weka to revision 12410 (post 3.7.13) to avoid performance bottleneck when using setOptions method fixed class SetupGenerator from module weka.core.classes added load_any_file method to the weka.core.converters module added save_any_file method to the weka.core.converters module if GridSearch instantiation (module weka.classifiers) fails, it now outputs message whether package installed and JVM with package support started	January 29, 2016, 05:22:58
0.3.4	added convenience method create_instances_from_lists to weka.core.dataset module to easily create an Instances object from numeric lists (x and y) added get_object_tags method to Tags class from module weka.core.classes, to allow obtaining weka.core.Tag array from the method of a JavaObject rather than a static field (MultiSearch) updated MultiSearch wrapper in module weka.classifiers to work with the multi-search package version 2016.1.15 or later	January 29, 2016, 05:21:27
0.3.3	updated to Weka 3.7.13 documentation now covers the API as well	September 26, 2015, 06:11:42
0.3.2	The "packages" parameter of the "weka.core.jvm.start()" function can be used for specifying an alternative Weka home directory now as well added "train_test_split" method to "weka.core.Instances" class to easily create train/test splits "evaluate_train_test_split" method of "weka.classifiers.Evaluation" class now uses the "train_test_split" method	June 28, 2015, 23:09:13
0.3.1	added "get_tags" class method to "Tags" class for easier instantiation of Tag arrays added "find" method to "Tags" class to locate "Tag" object that matches the string fixed "getitem" and "setitem" methods of the "Tags" class added "GridSearch" meta-classifier with convenience properties to module "weka.classifiers" added "SetupGenerator" and various parameter classes to "weka.core.classes" added "MultiSearch" meta-classifier with convenience properties to module "weka.classifiers" added "quote"/"unquote" and "backquote"/"unbackquote" methods to "weka.core.classes" module added "main" method to "weka.core.classes" for operations on options: join, split, code added support for option handling to "weka.core.classes" module	April 23, 2015, 00:06:57
0.3.0	added method "ndarray_to_instances" to "weka.converters" module for converting Numpy 2-dimensional array into "Instances" object added method "plot_learning_curve" to "weka.plot.classifiers" module for creating learning curves for multiple classifiers for a specific metric added plotting of experiments with "plot_experiment" methid in "weka.plot.experiments" module "Instance.create_instance" method now takes list of tuples (index, internal float value) when generating sparse instances added "weka.core.database" module for loading data from a database added "make_copy" class method to "Clusterer" class added "make_copy" class method to "Associator" class added "make_copy" class method to "Filter" class added "make_copy" class method to "DataGenerator" class most classes (like Classifier and Filter) now have a default classname value in the constructor added "TextDirectoryLoader" class to "weka.core.converters" moved all methods from "weka.core.utils" to "weka.core.classes" fixed "Attribute.index_of" method for determining label index fixed "Attribute.add_string_value" method (used incorrect JNI parameter) "create_instance" and "create_sparse_instance" methods of class "Instance" now ensure that list values are float added "to_help" method to "OptionHandler" class which outputs a help string generated from the base class's "globalInfo" and "listOptions" methods fixed "test_model" method of "Evaluation" class when supplying a "PredictionOutput" object (previously generated "No dataset structure provided!" exception) added "batch_finished" method to "Filter" class for incremental filtering added "line_plot" method to "weka.plot.dataset" module for plotting dataset using internal format (one line plot per instance) added "is_serializable" property to "JavaObject" class added "has_class" convenience property to "Instance" class added "repr" method to "JavaObject" classes (simply calls "toString()" method) added "Stemmer" class in module "weka.core.stemmers" added "Stopwords" class in module "weka.core.stopwords" added "Tokenizer" class in module "weka.core.tokenizers" added "StringToWordVector" filter class in module "weka.filters" added simple workflow engine (see documentation on Flow)	April 15, 2015, 12:37:22
0.2.2	added convenience methods "no_class" (to unset class) and "has_class" (class set?) to "Instances" class switched to using faster method objects for methods "classify_instance"/"distribution_for_instance" in "Classifier" class switched to using faster method objects for methods "cluster_instance"/"distribution_for_instance" in "Clusterer" class switched to using faster method objects for methods "class_index", "is_missing", "get/set_value", "get/set_string_value", "weight" in "Instance" class switched to using faster method objects for methods "input", "output", "outputformat" in "Filter" class switched to using faster method objects for methods "attribute", "attribute_by_name", "num_attributes", "num_instances", "class_index", "class_attribute", "set_instance", "get_instance", "add_instance" in "Instances" class	January 5, 2015, 03:43:56
0.2.1	added unit testing framework added method "refresh_cache()" to "weka/core/packages.py" to allow user to refresh local cache method "get_classname" in "weka.core.utils" now handles Python objects and class objects as well added convenience method "get_jclass" to "weka.core.utils" to instantiate a Java class added a "JavaArray" wrapper for "arrays, which allows getting/setting elements and iterating added property "classname" to class "JavaObject" for easy access to classname of underlying object added class method "parse_matlab" for parsing Matlab matrix strings to "CostMatrix" class "predictions" method of "Evaluation" class now return "None" if predictions are discarded "Associator.get_capabilities()" method is now a property: "Associator.capabilities" added wrapper classes for Java enums: "weka.core.classes.Enum" fixed retrieval of "sumSq" in "Stats" class (used by "AttributeStats") fixed "cluster_instance" method in "Clusterer" class fixed "filter" and "clusterer" properties in clusterer classes ("SingleClustererEnhancer", "FilteredClusterer") added "crossvalidate_model" method to "ClusterEvaluation" added "get_prc" method to "plot/classifiers.py" for calculating the area under the precision-recall curve "Filter.filter" method now handles list of "Instances" objects as well, applying the filter sequentially to all the datasets (allows generation of compatible train/test sets)	January 5, 2015, 00:30:01
0.2.0	NB: This release is not backwards compatible! requires "JavaBridge" 1.0.9 at least moved from Java-like get/set ("getIndex()" and "setIndex(int)") to nicer Python properties using Python properties (also only read-only ones) wherevere possible added "weka.core.version" for accessing the Weka version currently in use added "jwrapper" and "jclasswrapper" methods to "JavaObject" class (the mother of all objects in python-weka-wrapper) to allow generic access to an object's methods: http://pythonhosted.org//javabridge/highlevel.html#wrapping-java-objects-using-reflection added convenience methods "class_is_last()" and "class_is_first()" to "weka.core.Instances" class added convenience methods "delete_last_attribute()" and "delete_first_attribute()" to "weka.core.Instances" class	December 22, 2014, 09:21:53
0.1.17	fixed "setup.py" to download Weka 3.7.12 instead of 3.7.11 (this time correct URL)	December 17, 2014, 21:43:28
0.1.16	fixed setup.py to download Weka 3.7.12 instead of 3.7.11	December 17, 2014, 21:38:31
0.1.15	fixed "Instance.set_value" method: https://github.com/fracpete/python-weka-wrapper/issues/24 added sub-section "From source" to section on installing the library upgraded to Weka 3.7.12	December 17, 2014, 21:11:43
0.1.14	fixed setup.py to include the jars again when using eggs (via include_package_data etc) added detailed instructions for installing the library	December 16, 2014, 01:58:17
0.1.13	added "get_class" method to "weka.core.utils" which returns the Python class object associated with the classname in dot-notation "from_commandline" method in "weka.core.utils" now takes an optional "classname" argument, which is the classname (in dot-notation) of the wrapper class to return - instead of the generic "OptionHandler" added "Kernel" and "KernelClassifier" convenience classes to better handle kernel based classifiers	November 1, 2014, 09:59:54
0.1.12	added "create_string" class method to the "Attribute" class for creating a string attribute ROC/PRC curves can now consist of multiple plots (ie multiple class labels) switched command-line option handling from "getopt" to "argparse" fixed Instance.get_dataset(self) method added iterators for: rows/attributes in dataset, values in dataset row incremental loaders can be iterated now	October 17, 2014, 00:16:26
0.1.11	moved wekaexamples module to separate github project: https://github.com/fracpete/python-weka-wrapper-examples added "stratify", "train_cv" and "test_cv" methods to the Instances class fixed "to_summary" method of the Evaluation class: failed when providing a custom title	September 25, 2014, 00:39:02
0.1.10	fixed adding custom classpath using jvm.start(class_path=[...])	August 29, 2014, 05:00:14
0.1.9	added static methods to Instances class: summary, merge_instances, append_instances added methods to Instances class: delete_with_missing, equal_headers	August 29, 2014, 04:58:45
0.1.8	fixed installer: MANIFEST.in now includes CHANGES.rst and DESCRIPTION.rst as well	June 26, 2014, 02:38:12
0.1.7	fixed weka/plot/dataset.py imports to avoid error when testing for matplotlib availability Instance.create_instance (weka/core/dataset.py) now accepts Python list and Numpy array	June 26, 2014, 02:14:16
0.1.6	added troubleshooting section for Mac OSX to documentation recompiled helper jars with 1.6 rather than 1.7 added link to Google Group	May 29, 2014, 06:02:41
0.1.5	added CostMatrix support in the classifier evaluation fixed various retrievals of double arrays (accessed them incorrectly as float arrays), like distributionForInstance for a classifier Instances object can now retrieve all (internal) values of an attribute/column as numpy array added plotting of cluster assignments to weka.plot.clusterers fixed weka.core.utils.from_commandline method fixed weka.classifiers.PredictionOutput (get/set_header methods) predictions can be turned into an Instances object now using weka.classifiers.predictions_to_instances	May 23, 2014, 06:40:42
0.1.4	dependencies for plotting are now optional (pygraphviz, PIL, matplotlib) plots now support custom titles	May 19, 2014, 03:02:32
0.1.3	improved documentation added PRC curve plot aligned to PEP8 style guidelines fixed variety of little bugs (not so commonly used methods) fixed lib directory reference in make files for Java helper classes	May 17, 2014, 13:37:54
0.1.2	added matrix plot added scatter plot for two attributes fixes in constructors of classes added MultiFilter convenience class predictions (of classifiers) can now be collected and output using the PredictionOutput class added support for attribute statistics	May 13, 2014, 07:11:07
0.1.1	constructors now take list of commandline options as well added Weka package support (list/install/uninstall) ROC plotting for classifiers improved code documentation added more examples added more datasets using javabridge 1.0.1 now	May 2, 2014, 03:35:38
0.1.0	Initial Announcement on mloss.org.	April 28, 2014, 23:18:48

Comments

No one has posted any comments yet. Perhaps you'd like to be the first?

Leave a comment

You must be logged in to post comments.