require 'pycall'

pymath = PyCall.import_module('math')

pymath.pi

pymath.pi.class

pymath.sin(30 * Math::PI / 180)

Math.sin(30 * Math::PI / 180)

unless File.exist?('iris.data')
  system("curl -sfSLO https://archive.ics.uci.edu/ml/machine-learning-databases/iris/iris.data")
end

require 'pandas'

iris = Pandas.read_csv('iris.data', names: %w[SepalLength SepalWidth PetalLength PetalWidth Species])
nil

iris.head

iris.tail

iris.describe

# The last `.T` means transpose the dataframe.
# The reason why transpose the result is just the readability of the result.
iris.groupby('Species').describe.T

require 'matplotlib/iruby'
Matplotlib::IRuby.activate
Matplotlib::Pyplot.ioff

sns = PyCall.import_module('seaborn')

sns.pairplot(iris, hue: 'Species', markers: %w[o s D], diag_kws: {bins: 25})

iris['SepalRatio'] = iris['SepalLength'] / iris['SepalWidth']
iris['PetalRatio'] = iris['PetalLength'] / iris['PetalWidth']

iris[['SepalRatio', 'PetalRatio', 'Species']].groupby('Species').describe.T

sns.pairplot(iris, hue: 'Species', markers: %w[o s D], diag_kws: {bins: 25})

PCA = PyCall.import_module('sklearn.decomposition').PCA

pca = PCA.new
x_trans = pca.fit_transform(iris.iloc[0..-1, 0..3])
nil

iris_trans = Pandas::DataFrame.new(data: x_trans, columns: %w[X0 X1 X2 X3])
iris_trans['Species'] = iris['Species']
sns.pairplot(iris_trans, hue: 'Species', markers: %w[o s D], diag_kws: {bins: 25})

ModelSelection = PyCall.import_module('sklearn.model_selection')
x_train, x_test, y_train, y_test = ModelSelection.train_test_split(
  x_trans, iris['Species'],
  test_size: 0.4,
  random_state: 13
)
nil

SVC = PyCall.import_module('sklearn.svm').SVC
model = SVC.new

model.fit(x_train, y_train)
model.score(x_test, y_test)