clj-ml.data — tranchis/clj-ml 0.5.4

This namespace contains several functions for building creating and manipulating data sets and instances. The formats of these data sets as well as their classes can be modified and assigned to the instances. Finally data sets can be transformed into Clojure sequences that can be transformed using usual Clojure functions like map, reduce, etc.

attr-name clj

(attr-name attr)

attribute-at clj

(attribute-at dataset-or-instance index-or-name)

Returns attribute situated at the provided position or the provided name.

attribute-labels clj

(attribute-labels attr)

Returns the labels (possible values) for the given nominal attribute as keywords.

attribute-labels-as-strings clj

(attribute-labels-as-strings attr)

Returns the labels (possible values) for the given nominal attribute as strings.

attribute-labels-indexes clj

(attribute-labels-indexes attr)

Returns map of the labels (possible values) for the given nominal attribute as the keys with the values being the attributes index.

attribute-name-at clj

(attribute-name-at dataset-or-instance index-or-name)

Returns the name of an attribute situated at the provided position in the attributes definition of an instance or class

attribute-names clj

(attribute-names dataset-or-instance)

Returns the attribute names, as keywords, of the dataset or instance

attribute-value-fn clj

(attribute-value-fn ds attr-name)

Takes a dataset and an attribute name, returns a function that will select the attribute value of a given instance from the dataset.

attributes clj

(attributes dataset-or-instance)

Returns the attributes (weka.core.Attribute) of the dataset or instance

copy-dataset clj

(copy-dataset ds)

Uses the Instances constructor to copy a given dataset. Each Instance (row) will be shallow copied. So, while not all the data is copied you will be creating n new Instance objects, where n is the number of training examples.

dataset-add clj

(dataset-add dataset vector)

(dataset-add dataset weight vector)

Adds a new instance to a dataset. A clojure vector, map, or an Instance can be passed as arguments

dataset-append-name clj

(dataset-append-name dataset name-addition)

Sets the dataset's name

dataset-as-lists clj

(dataset-as-lists dataset)

Returns a lazy sequence of the dataset represented as lists. The values are the actual values (i.e. the string values) and not weka's internal double representation or clj-ml's keyword representation.

dataset-as-maps clj

(dataset-as-maps dataset)

Returns a lazy sequence of the dataset represetned as maps. This fn is preferale to mapping over a seq yourself with instance-to-map becuase it avoids redundant string interning of the attribute names.

dataset-as-vecs clj

(dataset-as-vecs dataset)

Returns a lazy sequence of the dataset represented as lists. The values are the actual values (i.e. the string values) and not weka's internal double representation or clj-ml's keyword representation.

dataset-at clj

(dataset-at dataset pos)

Returns the instance at a certain position from the dataset

dataset-attribute-at clj

(dataset-attribute-at dataset index-or-name)

dataset-attributes clj

(dataset-attributes dataset)

Returns the attributes (weka.core.Attribute) of the dataset or instance

dataset-class-index clj

(dataset-class-index dataset)

Returns the index of the class attribute for this dataset

dataset-class-labels clj

(dataset-class-labels dataset)

Returns the possible labels for the class attribute

dataset-class-name clj

(dataset-class-name dataset)

Returns the name of the class attribute in keyword form. Returns nil if not set.

dataset-class-values clj

(dataset-class-values dataset)

Returns a lazy-seq of the values for the dataset's class attribute. If the class is nominal then the string value (not keyword) is returned.

dataset-count clj

(dataset-count dataset)

Returns the number of elements in a dataset

dataset-extract-at clj

(dataset-extract-at dataset pos)

Removes and returns the instance at a certain position from the dataset

dataset-filename clj

(dataset-filename model-prefix model-dir tag)

dataset-format clj

(dataset-format dataset)

Returns the definition of the attributes of this dataset

dataset-index-attr clj

(dataset-index-attr dataset attr)

Returns the index of an attribute in the attributes definition of a dataset.

dataset-labels-at clj

(dataset-labels-at dataset-or-instance index-or-name)

dataset-name clj

(dataset-name dataset)

Returns the name of this dataset

dataset-nominal? clj

(dataset-nominal? dataset)

Returns boolean indicating if the class attribute is nominal

dataset-pop clj

(dataset-pop dataset)

Removes and returns the first instance in the dataset

dataset-remove-attribute-at clj

(dataset-remove-attribute-at dataset index)

Removes the attribute at the specified index

dataset-remove-class clj

(dataset-remove-class dataset)

Removes the class attribute from the dataset

dataset-replace-attribute! clj

(dataset-replace-attribute! dataset attr-name new-attr)

Replaces the specified attribute with the given one. (The attribute should be a weka.core.Attribute) This function only modifies the format of the dataset and does not deal with any instances. The intention is for this to be used on data-formats and not on datasets with data.

dataset-seq clj

(dataset-seq dataset)

Builds a new clojure sequence from this dataset

dataset-set-class clj

(dataset-set-class dataset index-or-name)

Sets the index of the attribute of the dataset that is the class of the dataset

dataset-set-name clj

(dataset-set-name dataset new-name)

Sets the dataset's name

dataset-weights clj

(dataset-weights dataset)

Returns a lazy-seq of the weights of the dataset instances.

do-split-dataset clj

(do-split-dataset ds & options)

The same as split-dataset but actual datasets are returned and not Delay objects that need dereffing.

docs-to-dataset clj

(docs-to-dataset docs model-prefix model-dir & opts)

Docs are expected to be maps with this structure: {:id [any], :has-class? [true/false], :title [string], :fulltext [string]}. Of course, title or fulltext could be nil. model-prefix is a filename prefix to saving/loading the model (necessary to initialize the string-to-wordvec filters), and model-dir is a folder to save/load the model.

opts are optional parameters: :keep-n [int], :lowercase [true/false], :words-to-keep [int], :normalize [int], :transform-tf [true/false], :transform-idf [true/false], :stemmer [true/false], :resample [true/false], :training [true/false], :testing [true/false].

A map is returned with structure {:dataset [the dataset], :docids [seq of docids as ordered in dataset]}.

enumeration-or-nil-seq clj

(enumeration-or-nil-seq s)

headers-only clj

(headers-only ds)

Returns a new weka dataset (Instances) with the same headers as the given one

instance-attribute-at clj

(instance-attribute-at instance index-or-name)

instance-attributes clj

(instance-attributes instance)

Returns the attributes (weka.core.Attribute) of the dataset or instance

instance-get-class clj

(instance-get-class instance)

Get the class attribute for this instance; returns nil if the class is "missing"

instance-index-attr clj

(instance-index-attr instance attr)

Returns the index of an attribute in the attributes definition of an instance or dataset

instance-set-class clj

(instance-set-class instance val)

Sets the value (label) of the class attribute for this instance

instance-set-class-missing clj

(instance-set-class-missing instance)

Sets the class to "missing"

instance-to-list clj

(instance-to-list instance)

Builds a list with the values of the instance

instance-to-map clj

(instance-to-map instance)

Builds a vector with the values of the instance

instance-to-vector clj

(instance-to-vector instance)

Builds a vector with the values of the instance

instance-value-at clj

(instance-value-at instance pos)

Returns the value of an instance attribute. A string, not a keyword is returned.

is-dataset? clj

(is-dataset? dataset)

Checks if the provided object is a dataset

is-instance? clj

(is-instance? instance)

Checks if the provided object is an instance

keyword-name clj

(keyword-name attr)

make-dataset clj

(make-dataset ds-name attributes capacity-or-labels & opts)

Creates a new dataset, empty or with the provided instances and options

make-instance clj

(make-instance dataset vector)

(make-instance dataset weight vector)

Creates a new dataset instance from a vector

nominal-attribute clj

(nominal-attribute attr-name labels)

Creates a nominal weka.core.Attribute with the given name and labels

nominal-attributes clj

(nominal-attributes dataset-or-instance)

Returns the string attributes (weka.core.Attribute) of the dataset or instance

numeric-attributes clj

(numeric-attributes dataset-or-instance)

Returns the numeric attributes (weka.core.Attribute) of the dataset or instance

randomize-dataset clj

(randomize-dataset ds)

(randomize-dataset ds seed)

Copies the given dataset and returns randomized version.

randomize-dataset! clj

(randomize-dataset! ds)

(randomize-dataset! ds seed)

Randomizes the dataset in place and returns the dataset. When no seed is provided then a randmon seed is created.

split-dataset clj

(split-dataset ds & [& {:keys [percentage num]}])

Splits the dataset into two parts based on either the ':percentage' given or the ':num' of instances. The first dataset returned will have 'percentage ammount of the original dataset and the second has the remaining portion. Both datasets are Delay objects that need to be dereffed. If you want to have the split immediately you can use do-split-dataset.

string-attributes clj

(string-attributes dataset-or-instance)

Returns the string attributes (weka.core.Attribute) of the dataset or instance

take-dataset clj

(take-dataset ds num)

Returns a subset of the given dataset containing the first 'num' instances.