skxcs 1.0

SciKit learn wrapper for XCS algorithm implementation.

skxcs

skxcs is a SciKit learn wrapper for implementation of XCS algorithm xcs.

Installation

Use the package manager pip to install skxcs. You need to have Cython installed.

pip install skxcs

Usage

Numeric Values

from skxcs.classifiers import XcsClassifier
import pandas as pd
from sklearn.model_selection import train_test_split

# Numeric values
numerical_frame = pd.read_csv('https://raw.githubusercontent.com/kliegr/arcBench/master/data/datasets/iris.csv')
numerical_frame.dropna(inplace=True)
y = numerical_frame['class']
numerical_frame.drop('class', axis=1, inplace=True)
X_train, X_test, y_train, y_test = train_test_split(numerical_frame, y, test_size=0.33)

classifier = XcsClassifier()

# If data input is non binary, classifier automatically uses MLDP discretizer for numeric values
# and one hot encoding for categorical values to transform data in both fit and predict methods.

classifier.fit(X_train, y_train)

# Get prediction array
y_pred = classifier.predict(X_test)

# Get pretty rules
for rule in classifier.get_pretty_rules():
    print(rule)

# To use get_pretty_rules or pretty_print_prediction methods,
# classifier has to transform train and test data first.

Categorical values

import pandas as pd
from skxcs.classifiers import XcsClassifier
from sklearn.model_selection import train_test_split

# Categorical values
categorical_frame = pd.read_csv('https://raw.githubusercontent.com/kliegr/arcBench/master/data/datasets/autos.csv')
categorical_frame.dropna(inplace=True)
y = categorical_frame['XClass']
categorical_frame = categorical_frame.select_dtypes(include=[object])
categorical_frame.drop('XClass', axis=1, inplace=True)
X_train, X_test, y_train, y_test = train_test_split(categorical_frame, y, test_size=0.25)
classifier = XcsClassifier()

# You can transform data yourself. You should either transform both training
# and testing data, or none of them. It is necessary to ensure correct values are passed to classifier.
X_train_bin = classifier.transform_df(X_train, y=y_train)
classifier.fit(X_train_bin, y_train)

# Note that we don't pass 'y' to transform method when we transform test data
X_test_bin = classifier.transform_df(X_test)

# pretty print prediction
result = classifier.pretty_print_prediction(X_test_bin)
print(result)

Contributing

...

License

MIT