require 'matplotlib/iruby'
Matplotlib::IRuby.activate
Matplotlib.interactive(false) # Interactive plotting OFF, necessary for inline plotting in IRuby

xs = Array.new(500) { 100 * rand }
ys = Array.new(500) { 100 * rand }
areas = Array.new(500) { 800 * rand }
nil

# scatter plot by seaborn
sns = PyCall.import_module('seaborn')
ax = sns.scatterplot(x: xs, y: ys, size: areas, hue: areas)
ax.set_xlabel('x')
ax.set_ylabel('y')
nil

system("wget https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass/pendigits")
system("wget https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass/pendigits.t")

require 'rumale'
samples, labels = Rumale::Dataset.load_libsvm_file('pendigits')

model = Rumale::Pipeline::Pipeline.new(steps: {
  rbf: Rumale::KernelApproximation::RBF.new(gamma: 0.0001, n_components: 800),
  svc: Rumale::LinearModel::SVC.new(reg_param: 0.0001, max_iter: 1000),
})

kf = Rumale::ModelSelection::StratifiedKFold.new(n_splits: 5, shuffle: true)
cv = Rumale::ModelSelection::CrossValidation.new(estimator: model, splitter: kf)

report = cv.perform(samples, labels)

mean_accuracy = report[:test_score].inject(:+) / kf.n_splits
puts("5-CV mean accuracy: %.1f %%" % (mean_accuracy * 100.0))

test_samples, test_labels = Rumale::Dataset.load_libsvm_file('pendigits.t')
accuracy = model.score(test_samples, test_labels)
puts("Accuracy: %.1f%%" % (100.0 * accuracy))