def test8(self, msg=warn("test ORF Regression")):
def f(x):
return math.sin(x[0]) if x[0] < x[1] else math.cos(x[1] +
math.pi / 2)
n = 1000
X = np.random.randn(n, 2)
y = map(f, X)
param = {
'minSamples': 10,
'minGain': 0,
'xrng': dataRange(X),
'maxDepth': 10
}
xtest = np.random.randn(n, 2)
ytest = map(f, xtest)
orf = ORF(param, numTrees=50)
for i in range(n):
orf.update(X[i, :], y[i])
preds = orf.predicts(xtest)
mse = mean(
map(lambda z: (z[0] - z[1]) * (z[0] - z[1]), zip(preds, ytest)))
print "ORF Regression:"
print "f(0,0): " + str(orf.predict([0, 0])) + " +/- " + str(
orf.predStat([0, 0], sd))
print "Mean size: " + str(orf.meanTreeSize())
print "SD size: " + str(orf.sdTreeSize())
print "Mean max depth: " + str(orf.meanMaxDepth())
print "SD max depth: " + str(orf.sdMaxDepth())
print "RMSE: " + str(math.sqrt(mse))
print
def test7(self, msg=warn("test ORF Classify")):
def f(x):
return int(x[0] * x[0] + x[1] * x[1] < 1)
n = 1000
X = np.random.randn(n, 2)
y = map(f, X)
param = {
'minSamples': 10,
'minGain': .01,
'numClasses': 2,
'xrng': dataRange(X)
}
orf = ORF(param, numTrees=50)
for i in range(n):
orf.update(X[i, :], y[i])
xtest = np.random.randn(n, 2)
ytest = map(f, xtest)
preds = orf.predicts(xtest)
conf = orf.confusion(xtest, ytest)
print
print sum(ytest)
orf.printConfusion(conf)
acc = map(lambda z: z[0] == z[1], zip(preds, ytest))
print "ORF Classify:"
print "Mean max depth: " + str(orf.meanMaxDepth())
print "Mean Size: " + str(orf.meanTreeSize())
print "SD Size: " + str(orf.sdTreeSize())
print "Accuracy: " + str(mean(acc))
print
def test5(self, msg=warn("test ORT Classify")):
def f(x):
return int(x[0] * x[0] + x[1] * x[1] < 1)
n = 1000
X = np.random.randn(n, 2)
y = map(f, X)
param = {
'minSamples': 10,
'minGain': .01,
'numClasses': 2,
'xrng': dataRange(X),
'maxDepth': 5
}
ort = ORT(param)
map(lambda i: ort.update(X[i, :], y[i]), range(n))
#ort.draw()
preds = map(lambda i: ort.predict(X[i, :]), range(n))
acc = map(lambda z: z[0] == z[1], zip(preds, y))
print "ORT Classify:"
print "Accuracy: " + str(mean(acc))
print "max depth: " + str(ort.tree.maxDepth())
print
def test6(self, msg=warn("test ORT Regression")):
def f(x):
return math.sin(x[0]) if x[0] < x[1] else math.cos(x[1] +
math.pi / 2)
n = 1000
X = np.random.randn(n, 2)
y = map(f, X)
param = {
'minSamples': 10,
'minGain': .01,
'xrng': dataRange(X),
'maxDepth': 5
}
ort = ORT(param)
for i in range(n):
ort.update(X[i, :], y[i])
#ort.draw()
preds = map(lambda i: ort.predict(X[i, :]), range(n))
mse = mean(map(lambda z: (z[0] - z[1]) * (z[0] - z[1]), zip(preds, y)))
print "ORT Regression:"
print "RMSE: " + str(math.sqrt(mse))
print "max depth: " + str(ort.tree.maxDepth())
print
import numpy as np
import math
from ORFpy import ORF, dataRange
def g(x):
return math.sin(x[0]) if x[0] < x[1] else math.cos(x[1] + math.pi / 2)
n = 10
X = np.random.randn(n, 2)
y = map(g, X)
param = {'minSamples': 10, 'minGain': 0, 'xrng': dataRange(X), 'maxDepth': 10}
xtest = np.random.randn(n, 2)
ytest = map(g, xtest)
orf = ORF(param, numTrees=50)
for i in range(n):
orf.update(X[i, :], y[i])
preds = orf.predicts(xtest)
sse = sum(map(lambda z: (z[0] - z[1]) * (z[0] - z[1]), zip(preds, ytest)))
rmse = math.sqrt(sse / float(len(preds)))
print "RMSE: " + str(round(rmse, 2))
# RMSE: 0.22
print "Online train finish!"
def save(self):
# 保存Model(注:save文件夹要预先建立，否则会报错)
joblib.dump(self.orf, "./save/rf.pkl")
def load(self):
# 读取Model
self.orf = joblib.load("./save/rf.pkl")
def test(self):
preds = self.orf.predicts(self.testX)
# RMSE 均方根误差亦称标准误差
sse = sum(map(lambda z: (z[0] - z[1]) * (z[0] - z[1]), zip(preds, self.testY)))
rmse = math.sqrt(sse / float(len(preds)))
print "RMSE: " + str(round(rmse, 2)) + "\n"
if __name__ == "__main__":
fopen = open("data/data.csv", 'r')
tempData = []
for eachLine in fopen:
eachLineData = eachLine.split(",")
eachLineData = np.array(map(lambda x: float(x), eachLineData))
tempData.append(eachLineData)
tempData = np.array(tempData)
fopen.close()
param = {'minSamples': 10, 'minGain': 0, 'xrng': dataRange(tempData[:,:722]), 'maxDepth': 10}
orf = OnlineRandomForest(param=param,numTrees=50,input=722)
orf.offline_train()
orf.online_train()
from ORF import OnlineRandomForest
from ORFpy import dataRange
import numpy as np
if __name__ == "__main__":
fopen = open("data/data.csv", 'r')
tempData = []
for eachLine in fopen:
eachLineData = eachLine.split(",")
eachLineData = np.array(map(lambda x: float(x), eachLineData))
tempData.append(eachLineData)
tempData = np.array(tempData)
fopen.close()
param = {
'minSamples': 10,
'minGain': 0,
'xrng': dataRange(tempData[:, :722]),
'maxDepth': 10
}
orf = OnlineRandomForest(param=param, numTrees=50, input=722)
orf.online_train()