Write a Python algorithm class

To add a custom algorithm to the Splunk Machine Learning Toolkit (MLTK) you must write a Python algorithm class. The algorithm class must implement certain methods to properly operate with upstream processes. These methods are the entry points to an algorithm, where the data and options are specified as arguments.

Best practices

Follow these best practices when writing an algorithm class:

• Assume invalid input.
• If there is a parameter passed in make sure you check that it is valid.
• If you require a particular field, for example, _time, make sure you check for its presence and error accordingly.

Methods

Methods are the entry point to the custom algorithm. Refer to the following table for details about each method:

Method	Required	Arguments
__init__	Yes	self, options
fit	Yes	self, df, options
apply	Only for saved models	self, df, options
register_codecs	Only for saved models	(none)
partial_fit	No	self, df, options
summary	No	self, options

Arguments

Specify data and options as arguments. Refer to the following table for details about each argument:

Argument	Description
options	Options include: args (list): A list of the fields used. params (dict): Any parameters (key-value) pairs in the search. feature_variables (list): The fields to be used as features. target_variable (list): The target field for prediction. algo_name (str): The name of algorithm. mlspl_limits (dict): mlspl.conf stanza properties that may be used in utility methods. Other options that may exist depending on the search: model_name (str): The name of the model being saved (into clause). output_name (str): The name of the output (as clause).
df	A pandas DataFrame of the input data from the search results. See http://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.html.

Example:

The following example is a dictionary of information from the search:

{
 	 'args': [u'sepal_width', u'petal*'],
	 'params': {u'fit_intercept': u't'},
	 'feature_variables': ['petal*'],
	 'target_variable': ['sepal_width']
	 'algo_name': u'LinearRegression',
         'mlspl_limits': { ... },
 }

Attributes

Inside of the fit method, two attributes can be attached to self by the search command. Refer to the following table for details about each attribute:

Attribute	Description
self.feature_variables (list)	The wildcard matched list of fields present from the search
self.target_variable (str)	The name of the target field. This field is only present if the from clause is used.