×1
i_list = []
2
score_list = []
3
for i in range(1,100,1):
4i=i/1.
5clf_tree = DecisionTreeClassifier(max_depth =i )   #使用SVM分类器来判断涨跌
6clf_tree.fit(X_train, y_train)
7i_list.append(i)
8score_list.append(clf_tree.score(X_test, y_test))
9
​
10
score_list_df =  pd.DataFrame({'max_depth':i_list,'score_list':score_list})
11
score_list_df.plot(x='max_depth' ,y='score_list',title='score change with max_depth')

然后是min_samples_leaf值，同理。这里是从0.1到10变动范围

i_list = []
score_list = []
for i in range(1,100,1):i=i/10.clf_tree = DecisionTreeClassifier(min_samples_leaf  = i ) clf_tree.fit(X_train, y_train)i_list.append(i)score_list.append(clf_tree.score(X_test, y_test))score_list_df =  pd.DataFrame({'min_samples_leaf':i_list,'score_list':score_list})
score_list_df.plot(x='min_samples_leaf' ,y='score_list',title='score change with min_samples_leaf')

然后是min_samples，这里选用2~50

i_list = []
score_list = []min_samples  =  range(2,50,1)
for i in min_samples :clf_tree = DecisionTreeClassifier(min_samples_split  = i ) clf_tree.fit(X_train, y_train)i_list.append(i)score_list.append(clf_tree.score(X_test, y_test))score_list_df =  pd.DataFrame({'min_samples_split':i_list,'score_list':score_list})
score_list_df.plot(x='min_samples_split' ,y='score_list',title='score change with min_samples_split')

i_list = []
score_list = []min_weight  =  [x /1000. for x in range(1,500,1)]
for i in min_weight :clf_tree = DecisionTreeClassifier(min_weight_fraction_leaf  = i ) clf_tree.fit(X_train, y_train)i_list.append(i)score_list.append(clf_tree.score(X_test, y_test))score_list_df =  pd.DataFrame({'min_weight_fraction_leaf':i_list,'score_list':score_list})
score_list_df.plot(x='min_weight_fraction_leaf' ,y='score_list',title='score change with min_weight')

知道了大致参数最优范围以后，我们使用grisearchCV在这个范围内找到最优解。

from sklearn.grid_search import GridSearchCV
from sklearn.cross_validation import ShuffleSplit
params = {'max_depth':range(1,50,1),'min_samples_split':range(2,50,1)}X_train,X_test, y_train, y_test = cross_validation.train_test_split(features_scaler, lables, test_size = 0.2, random_state = 0)clf_tree = DecisionTreeClassifier() # cv_sets = ShuffleSplit(X_train.shape[0], n_iter = 10, test_size = 0.20, random_state = 0)grid = GridSearchCV(clf_tree, params)
grid = grid.fit(X_train, y_train)
print grid.best_estimator_

然后在最优解的基础上再次计算一次准确率

from sklearn import cross_validation
from sklearn import preprocessing
scaler = preprocessing.StandardScaler().fit(features)
features_scaler = scaler.transform(features)
#上面两行代码用来标准化数据X_train,X_test, y_train, y_test = cross_validation.train_test_split(features_scaler, lables, test_size = 0.2, random_state = 0)clf_tree = DecisionTreeClassifier(max_depth = 32 , min_samples_split=3 )
clf_tree.fit(X_train, y_train)
print "预测准确率为：%0.2f" % (clf_tree.score(X_test, y_test))

然后我们通过返回数据重要性，来看看对我们决策树哪些特征影响最大。

通过下方的表格（和直方图）发现影响最大的特征主要是30日的最小值，30日的最大值以及30日的最大日波动，然后是当日交易量。可以简单认为，前30日的波动幅度，最大震荡能一定程度上反映出未来30日后的市场涨跌趋势。

为了判断模型是否稳健，我们让训练集合处于变化中，然后观察随着训练集合的变化，准确率的波动范围图。这里采取的是1000~2500数据每10个变化一次。

从图的表现可以看出，波动范围0.1左右，平均线在0.88左右，大致还算稳健……吧。

num_list = []
score_list = []
for i in xrange((len(features_scaler)-1000)/2):num_now = len(features_scaler)%2 + 2*i +1000X_train,X_test, y_train, y_test = cross_validation.train_test_split(features_scaler[:num_now], lables[:num_now], test_size = 0.2, random_state = 0)clf_tree = DecisionTreeClassifier( max_depth = 32 , min_samples_split=3 )   #使用SVM分类器来判断涨跌clf_tree.fit(X_train, y_train)num_list.append(num_now)score_list.append(clf_tree.score(X_test, y_test))score_list_df =  pd.DataFrame({'sets_num':num_list,'accuracy':score_list})
score_list_df.plot(x='sets_num' ,y='accuracy',title='Accuracy with sets')

import random
2
start = '2016-01-01'                       # 回测起始时间
3
end = '2016-12-15'                         # 回测结束时间
4
benchmark = 'HS300'                        # 策略参考标准
5
universe = set_universe('HS300')  # 证券池，支持股票和基金
6
capital_base = 100000                      # 起始资金
7
freq = 'd'                                 # 策略类型，'d'表示日间策略使用日线回测，'m'表示日内策略使用分钟线回测
8
refresh_rate = 1                           # 调仓频率，表示执行handle_data的时间间隔，若freq = 'd'时间间隔的单位为交易日，若freq = 'm'时间间隔为分钟
9
stock = '000300' #预测的指数，沪深300指数。和策略参考池一致。
10
fields = ['tradeDate','closeIndex', 'highestIndex','lowestIndex', 'turnoverVol','CHG','CHGPct']
11
#tradeDate是交易日、closeIndex是收盘指数、highestIndex是当日最大指数，lowestIndex是当日最小指数，CHG是涨跌
12
​
13
def initialize(account):                   # 初始化虚拟账户状态
14pass
15
​
16
features_list = []
17
def handle_data(account):
18# 生成买入列表
19last_date = account.previous_date.strftime("%Y-%m-%d") #获取上一个交易日日期并格式化
20begin_date = pd.date_range(end=last_date,periods=60)[0] #获取60日之前的交易日日期
21begin_date = begin_date.strftime("%Y-%m-%d") #格式化这个日期
22to_class = DataAPI.MktIdxdGet(ticker='000300',beginDate=begin_date,endDate=last_date,field=fields,pandas="1")
23to_class = to_class.dropna()
24to_class = to_class[-30:] #获取我们要的30天的指数信息
25to_class_date = to_class.set_index('tradeDate')
26to_class_date['max_difference'] = to_class_date['highestIndex'] - to_class_date['lowestIndex']
2728to_class_date_max_of_30day = max(to_class_date['highestIndex'])
29#找出前30天最大值。
30to_class_date_min_of_30day = min(to_class_date['lowestIndex'])
31#找出前30天最小值
32to_class_date_max_difference_of_30day = max(to_class_date['max_difference'])
33#找出前30天最大日波动
3435features_for_predict = to_class_date[-1:]
36features_for_predict['max_of_30day'] = to_class_date_max_of_30day
37features_for_predict['min_of_30day'] = to_class_date_min_of_30day
38features_for_predict['max_difference_of_30day'] = to_class_date_max_difference_of_30day
3940features_fp_scaler = scaler.transform(features_for_predict)
41predict_up = clf_tree.predict(features_fp_scaler)
42
​
43#预测30天后的收盘是涨还是跌。
44random.shuffle(account.universe)
45for stock in account.universe:         # 股票是股票池中的股票，并且优矿帮你自动剔除了当天停牌退市的股票
46p = account.reference_price[stock]        # 股票前一天的收盘价
47cost = account.security_cost.get(stock)  # 股票的平均持仓成本
48if predict_up and not cost:                        # 判断当前没有买入该股票
49order_pct_to(stock, 0.10)          # 将满足条件的股票买入，总价值占虚拟帐户的10%
50elif cost and p >= cost * 1.20:        # 卖出条件，当p这个价格涨幅到买入价的1.20倍；
51order_to(stock, 0)        # 将满足条件的股票卖到剩余0股，即全部卖出

×1
import random
2
start = '2015-01-01'                       # 回测起始时间
3
end = '2016-12-15'                         # 回测结束时间
4
benchmark = 'HS300'                        # 策略参考标准
5
universe = set_universe('HS300')  # 证券池，支持股票和基金
6
capital_base = 100000                      # 起始资金
7
freq = 'd'                                 # 策略类型，'d'表示日间策略使用日线回测，'m'表示日内策略使用分钟线回测
8
refresh_rate = 1                           # 调仓频率，表示执行handle_data的时间间隔，若freq = 'd'时间间隔的单位为交易日，若freq = 'm'时间间隔为分钟
9
stock = '000300' #预测的指数，沪深300指数。和策略参考池一致。
10
fields = ['tradeDate','closeIndex', 'highestIndex','lowestIndex', 'turnoverVol','CHG','CHGPct']
11
#tradeDate是交易日、closeIndex是收盘指数、highestIndex是当日最大指数，lowestIndex是当日最小指数，CHG是涨跌
12
​
13
def initialize(account):                   # 初始化虚拟账户状态
14pass
15
​
16
features_list = []
17
def handle_data(account):
18# 生成买入列表
19last_date = account.previous_date.strftime("%Y-%m-%d") #获取上一个交易日日期并格式化
20begin_date = pd.date_range(end=last_date,periods=60)[0] #获取60日之前的交易日日期
21begin_date = begin_date.strftime("%Y-%m-%d") #格式化这个日期
22to_class = DataAPI.MktIdxdGet(ticker='000300',beginDate=begin_date,endDate=last_date,field=fields,pandas="1")
23to_class = to_class.dropna()
24to_class = to_class[-30:] #获取我们要的30天的指数信息
25to_class_date = to_class.set_index('tradeDate')
26to_class_date['max_difference'] = to_class_date['highestIndex'] - to_class_date['lowestIndex']
2728to_class_date_max_of_30day = max(to_class_date['highestIndex'])
29#找出前30天最大值。
30to_class_date_min_of_30day = min(to_class_date['lowestIndex'])
31#找出前30天最小值
32to_class_date_max_difference_of_30day = max(to_class_date['max_difference'])
33#找出前30天最大日波动
3435features_for_predict = to_class_date[-1:]
36features_for_predict['max_of_30day'] = to_class_date_max_of_30day
37features_for_predict['min_of_30day'] = to_class_date_min_of_30day
38features_for_predict['max_difference_of_30day'] = to_class_date_max_difference_of_30day
3940features_fp_scaler = scaler.transform(features_for_predict)
41predict_up = clf_tree.predict(features_fp_scaler)
42
​
43#预测30天后的收盘是涨还是跌。
44random.shuffle(account.universe)
45for stock in account.universe:         # 股票是股票池中的股票，并且优矿帮你自动剔除了当天停牌退市的股票
46p = account.reference_price[stock]        # 股票前一天的收盘价
47cost = account.security_cost.get(stock)  # 股票的平均持仓成本
48if p<4 and  not cost:                        # 判断当前没有买入该股票
49order_pct_to(stock, 0.10)          # 将满足条件的股票买入，总价值占虚拟帐户的10%
50elif cost and p >= cost * 1.20:        # 卖出条件，当p这个价格涨幅到买入价的1.20倍；
51order_to(stock, 0)        # 将满足条件的股票卖到剩余0股，即全部卖出