交叉验证分类器错误泡沫

# Note: install jfreechartOffscreenRenderer package as well for JFreeChart library

# imports
import weka.classifiers.Evaluation as Evaluation
import weka.classifiers.functions.LinearRegression as LinearRegression
import weka.core.converters.ConverterUtils.DataSource as DS
import java.util.Random as Random
import org.jfree.data.xy.DefaultXYZDataset as DefaultXYZDataset
import org.jfree.chart.ChartFactory as ChartFactory
import org.jfree.chart.plot.PlotOrientation as PlotOrientation
import org.jfree.chart.ChartPanel as ChartPanel
import org.jfree.chart.renderer.xy.XYBubbleRenderer as XYBubbleRenderer
import org.jfree.chart.ChartUtilities as ChartUtilities
import javax.swing.JFrame as JFrame
import java.io.File as File
import os

# load data
data = DS.read(os.environ.get("MOOC_DATA") + os.sep + "bodyfat.arff")
data.setClassIndex(data.numAttributes() - 1)

# configure classifier
cls = LinearRegression()
cls.setOptions(["-C", "-S", "1"])

# cross-validate classifier
evl = Evaluation(data)
evl.crossValidateModel(cls, data, 10, Random(1))

# collect predictions
act = []
prd = []
err = []
for i in range(evl.predictions().size()):
    prediction = evl.predictions().get(i)
    act.append(prediction.actual())
    prd.append(prediction.predicted())
    err.append(abs(prediction.actual() - prediction.predicted()))
    
# create plot
plotdata = DefaultXYZDataset()
plotdata.addSeries("LR on " + data.relationName(), [act, prd, err])
plot = ChartFactory.createScatterPlot(\
    "Classifier errors", "Actual", "Predicted", \
    plotdata, PlotOrientation.VERTICAL, True, True, True)
plot.getPlot().setRenderer(XYBubbleRenderer())

# display plot
frame = JFrame()
frame.setTitle("Weka")
frame.setSize(800, 800)
frame.setLocationRelativeTo(None)
frame.getContentPane().add(ChartPanel(plot))
frame.setVisible(True)