# -*- coding: utf-8 -*-
"""
@author: Satoshi Hara
"""
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import matplotlib.cm as cm
import matplotlib.colorbar as colorbar
import pylab as pl
def plotRule(mdl, X, d1, d2, alpha=0.8, filename='', rnum=-1, plot_line=[]):
cmap = cm.get_cmap('cool')
fig, (ax1, ax2) = plt.subplots(1, 2, gridspec_kw = {'width_ratios':[19, 1]})
if rnum <= 0:
rnum = len(mdl.rule_)
else:
rnum = min(len(mdl.rule_), rnum)
idx = np.argsort(mdl.weight_[:rnum])
for i in range(rnum):
r = mdl.rule_[idx[i]]
box, vmin, vmax = __r2boxWithX(r, X)
if mdl.modeltype_ == 'regression':
c = cmap(mdl.pred_[idx[i]])
elif mdl.modeltype_ == 'classification':
r = mdl.pred_[idx[i]] / max(np.unique(mdl.pred_).size - 1, 1)
c = cmap(r)
ax1.add_patch(pl.Rectangle(xy=[box[0, d1], box[0, d2]], width=(box[1, d1] - box[0, d1]), height=(box[1, d2] - box[0, d2]), facecolor=c, linewidth='2.0', alpha=alpha))
if len(plot_line) > 0:
for l in plot_line:
ax1.plot(l[0], l[1], 'k--')
ax1.set_xlabel('x1', size=22)
ax1.set_ylabel('x2', size=22)
ax1.set_title('Simplified Model (K = %d)' % (rnum,), size=28)
colorbar.ColorbarBase(ax2, cmap=cmap, format='%.1f')
ax2.set_ylabel('Predictor y', size=22)
plt.show()
if not filename == '':
plt.savefig(filename, format="pdf", bbox_inches="tight")
plt.close()
def plotEachRule(mdl, X, d1, d2, alpha=0.8, filename='', rnum=-1, plot_line=[]):
if rnum <= 0:
rnum = len(mdl.rule_)
else:
rnum = min(len(mdl.rule_), rnum)
m = rnum // 4
if m * 4 < rnum:
m += 1
cmap = cm.get_cmap('cool')
fig, ax = plt.subplots(m, 4 + 1, figsize=(4 * 4, 3 * m), gridspec_kw = {'width_ratios':[15, 15, 15, 15, 1]})
idx = np.argsort(mdl.weight_[:rnum])
for i in range(rnum):
j = i // 4
k = i - 4 * j
r = mdl.rule_[idx[i]]
box, vmin, vmax = __r2boxWithX(r, X)
if mdl.modeltype_ == 'regression':
c = cmap(mdl.pred_[idx[i]])
elif mdl.modeltype_ == 'classification':
r = mdl.pred_[idx[i]] / max(np.unique(mdl.pred_).size - 1, 1)
c = cmap(r)
ax[j, k].add_patch(pl.Rectangle(xy=[box[0, d1], box[0, d2]], width=(box[1, d1] - box[0, d1]), height=(box[1, d2] - box[0, d2]), facecolor=c, linewidth='2.0', alpha=alpha))
if len(plot_line) > 0:
for l in plot_line:
ax[j, k].plot(l[0], l[1], 'k--')
ax[j, k].set_xlim([0, 1])
ax[j, k].set_ylim([0, 1])
if k == 3:
cbar = colorbar.ColorbarBase(ax[j, -1], cmap=cmap, format='%.1f', ticks=[0.0, 0.5, 1.0])
cbar.ax.set_yticklabels([0.0, 0.5, 1.0])
ax[j, -1].set_ylabel('Predictor y', size=12)
plt.show()
if not filename == '':
plt.savefig(filename, format="pdf", bbox_inches="tight")
plt.close()
def __r2boxWithX(r, X):
vmin = np.min(X, axis=0)
vmax = np.max(X, axis=0)
box = np.c_[vmin, vmax].T
for rr in r:
if rr[1] == 0:
box[1, rr[0]-1] = np.minimum(box[1, rr[0]-1], rr[2])
else:
box[0, rr[0]-1] = np.maximum(box[0, rr[0]-1], rr[2])
return box, vmin, vmax
