def grid_search(y,x,fw,c_range,g_range):
'''c,g从的变化
c_begin, c_end, c_step = -5, 15, 2
g_begin, g_end, g_step = 3, -15, -2
'''
best_rate,best_c,best_g = 0.0,2**-5,2**-15
esp = 0.001*100 #如果rate的增长幅度不超过esp，且c比best_c大时，就不会更新best_c
job = calculate_jobs(permute_sequence(c_range),permute_sequence(g_range))
fw.write("需要试验的组合有%s个：\n"%(len(job)))
for (c1,g1) in job:
fw.write(str(2**c1)+"\t"+str(2**g1)+"\n")
fw.write("下面是各种组合得到的交叉验证的效果：\n")
for (c1,g1) in job:
c,g=2**c1,2**g1
param="-v 5 "
if len(c_range)>2:
param+=" -c "+str(c)
if len(g_range)>2:
param+=" -g "+str(g)
rate = tms_svm.train(y,x,param)
if (c < best_c and rate > best_rate-esp ) or (c > best_c and rate-esp > best_rate) or (c== best_c and rate > best_rate) or (abs(rate-best_rate)<esp and g==best_g and c<best_c):
best_rate = rate
best_c,best_g=c,g
fw.write(time.strftime('%Y-%m-%d %H:%M:%S',time.localtime(time.time()))+"\t")
fw.write("%s\t%s\t%s\t(best c = %s,g = %s,rate =%s)\n" %(c,g,rate,best_c,best_g,best_rate))
fw.flush()
print "%s %s %s (best c = %s,g = %s,rate =%s)\n" %(c,g,rate,best_c,best_g,best_rate)
fw.write("(best c = %s,g = %s,rate =%s)\n" %(best_c,best_g,best_rate))
return best_c,best_g
def reducer(self, key, values):
'''reducer function'''
prob_y = []
prob_x = []
line = key.split(None, 1)
#设置SVM训练的参数
if sum([1 for i in line]) == 1:
svm_param = " -v 5 -c " + str(line[0])
else:
if sum([1 for i in line]) >= 2:
svm_param = " -v 5 -c " + str(line[0]) + " -g " + str(line[1])
#对训练样本进行汇总整理
for value in values:
value = value.split(None, 1)
if len(value) == 1: value += ['']
label, features = value
xi = {}
for e in features.split():
ind, val = e.split(":")
xi[int(ind)] = float(val)
prob_y += [float(label)]
prob_x += [xi]
#对得到的参数与训练样本进行训练
tms_svm.set_svm_type(svm_type)
ratio = tms_svm.train(prob_y, prob_x, svm_param)
self.write_output(key, str(ratio))
def reducer(self, key, values):
'''reducer function'''
prob_y=[]
prob_x=[]
line=key.split(None,1)
#设置SVM训练的参数
if sum([1 for i in line])==1:
svm_param = " -v 5 -c "+str(line[0])
else :
if sum([1 for i in line])>=2:
svm_param = " -v 5 -c "+str(line[0])+" -g "+str(line[1])
#对训练样本进行汇总整理
for value in values:
value = value.split(None,1)
if len(value)==1: value+=['']
label, features = value
xi={}
for e in features.split():
ind, val = e.split(":")
xi[int(ind)] = float(val)
prob_y +=[float(label)]
prob_x +=[xi]
#对得到的参数与训练样本进行训练
tms_svm.set_svm_type(svm_type)
ratio = tms_svm.train(prob_y,prob_x,svm_param)
self.write_output( key, str(ratio))
def ctm_train_model(sample_save_path,svm_type,param,model_save_path):
'''训练模型，输入样本文件，训练的参数，模型的保存地址，最后会给出模型在训练样本上的测试结果。'''
tms_svm.set_svm_type(svm_type)
y,x = tms_svm.read_problem(sample_save_path)
m = tms_svm.train(y,x,param)
tms_svm.save_model(model_save_path,m)
labels = {}.fromkeys(y).keys()
if len(labels)>2:
pred_labels, (Micro, Macro, ACC), pred_values = tms_svm.predict(y,x,m)
print "(Micro=%g, Macro=%g, ACC=%g)"%(Micro, Macro, ACC)
else:
pred_labels, (f_score,recall,presion), pred_values=tms_svm.predict(y,x,m)
print "(f_score=%g,recall=%g,presion=%g)"%(f_score,recall,presion)
return m
def grid_search(y, x, fw, c_range, g_range):
'''c,g从的变化
c_begin, c_end, c_step = -5, 15, 2
g_begin, g_end, g_step = 3, -15, -2
'''
best_rate, best_c, best_g = 0.0, 2**-5, 2**-15
esp = 0.001 * 100 #如果rate的增长幅度不超过esp，且c比best_c大时，就不会更新best_c
job = calculate_jobs(permute_sequence(c_range), permute_sequence(g_range))
fw.write("需要试验的组合有%s个：\n" % (len(job)))
for (c1, g1) in job:
fw.write(str(2**c1) + "\t" + str(2**g1) + "\n")
fw.write("下面是各种组合得到的交叉验证的效果：\n")
for (c1, g1) in job:
c, g = 2**c1, 2**g1
param = "-v 5 "
if len(c_range) > 2:
param += " -c " + str(c)
if len(g_range) > 2:
param += " -g " + str(g)
rate = tms_svm.train(y, x, param)
if (c < best_c and rate > best_rate - esp) or (
c > best_c and rate - esp > best_rate) or (
c == best_c
and rate > best_rate) or (abs(rate - best_rate) < esp
and g == best_g and c < best_c):
best_rate = rate
best_c, best_g = c, g
fw.write(
time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())) +
"\t")
fw.write("%s\t%s\t%s\t(best c = %s,g = %s,rate =%s)\n" %
(c, g, rate, best_c, best_g, best_rate))
fw.flush()
print "%s %s %s (best c = %s,g = %s,rate =%s)\n" % (c, g, rate, best_c,
best_g, best_rate)
fw.write("(best c = %s,g = %s,rate =%s)\n" % (best_c, best_g, best_rate))
return best_c, best_g
