商业数据分析--股票涨跌预测模型搭建

姜智浩 Lv5
  2025-09-25 11:45:14 2025-09-25 11:45:14 Created   2025-09-25 09:44:49 2025-09-25 09:44:49 Updated      2k Words  9 Mins  

声明

本文代码均保存在
https://github.com/super-213/business_data_analysis
有需要的可以自行下载

导入库

1
2
3
4
5
6
7
8
9
import tushare as ts  # 股票基本数据相关库
import numpy as np  # 科学计算相关库
import pandas as pd  # 科学计算相关库  
import talib  # 股票衍生变量数据相关库
import matplotlib.pyplot as plt  # 引入绘图相关库，可视化
from sklearn.ensemble import RandomForestClassifier  # 引入分类决策树模型
from sklearn.metrics import accuracy_score  # 引入准确度评分函数
import warnings
warnings.filterwarnings("ignore") # 忽略警告信息，警告非报错，不影响代码执行
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
# 1.股票基本数据获取
df = ts.get_k_data('000002',start='2015-01-01',end='2019-12-31')
df = df.set_index('date')  # 设置日期为索引

# 2.简单衍生变量构造
df['close-open'] = (df['close'] - df['open'])/df['open']
df['high-low'] = (df['high'] - df['low'])/df['low']

df['pre_close'] = df['close'].shift(1)  # 该列所有往下移一行形成昨日收盘价
df['price_change'] = df['close']-df['pre_close']
df['p_change'] = (df['close']-df['pre_close'])/df['pre_close']*100

# 3.移动平均线相关数据构造
df['MA5'] = df['close'].rolling(5).mean()
df['MA10'] = df['close'].rolling(10).mean()
df.dropna(inplace=True)  # 删除空值

# 4.通过Ta_lib库构造衍生变量
df['RSI'] = talib.RSI(df['close'], timeperiod=12)  # 相对强弱指标
df['MOM'] = talib.MOM(df['close'], timeperiod=5)  # 动量指标
df['EMA12'] = talib.EMA(df['close'], timeperiod=12)  # 12日指数移动平均线
df['EMA26'] = talib.EMA(df['close'], timeperiod=26)  # 26日指数移动平均线
df['MACD'], df['MACDsignal'], df['MACDhist'] = talib.MACD(df['close'], fastperiod=12, slowperiod=26, signalperiod=9)  # MACD值
df.dropna(inplace=True)  # 删除空值

库太老了 我使用其他数据获取库
用其他更现代的库来获取股票数据

akshare

pip install akshare

查看数据

1
2
3
4
5
import akshare as ak
df = ak.stock_zh_a_hist(symbol="000002", period="daily", start_date="20150101", end_date="20191231", adjust="")
df = df.set_index('日期')

df.head()
日期 股票代码 开盘 收盘 最高 最低 成交量 成交额 振幅 涨跌幅 涨跌额 换手率
2015-01-05 000002 14.39 14.91 15.29 14.22 6560836 9.700712e+09 7.70 7.27 1.01 6.76
2015-01-06 000002 14.60 14.36 14.99 14.05 3346347 4.839616e+09 6.30 -3.69 -0.55 3.45
2015-01-07 000002 14.26 14.23 14.50 14.00 2642051 3.772151e+09 3.48 -0.91 -0.13 2.72
2015-01-08 000002 14.32 13.59 14.37 13.46 2639394 3.629554e+09 6.39 -4.50 -0.64 2.72
2015-01-09 000002 13.54 13.45 14.22 13.29 3294584 4.521978e+09 6.84 -1.03 -0.14 3.39

画k线图

针对股票市场 我们肯定要查看其k线图

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
import plotly.graph_objects as go

data = df.loc["2015-01-01":"2015-01-31"]

fig = go.Figure(data=[go.Candlestick(
    x=data.index,
    open=data["Open"],
    high=data["High"],
    low=data["Low"],
    close=data["Close"],
    increasing_line_color="red",
    decreasing_line_color="green"
)])

fig.update_layout(
    title="000002 万科A K线图 (2015-01)",
    xaxis_title="日期",
    yaxis_title="价格",
    xaxis_rangeslider_visible=False
)

fig.show()

photo

随机森林

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report, accuracy_score

df = ak.stock_zh_a_hist(symbol="000002", period="daily", start_date="20150101", end_date="20150131", adjust="qfq")

df["label"] = (df["收盘"].shift(-1) > df["收盘"]).astype(int)  # 次日涨=1，跌=0

X = df[["开盘", "最高", "最低", "收盘", "成交量", "成交额", "振幅", "涨跌幅", "换手率"]]
y = df["label"]

X = X[:-1]
y = y[:-1]

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, shuffle=False)

rf = RandomForestClassifier(n_estimators=200, random_state=42)
rf.fit(X_train, y_train)

y_pred = rf.predict(X_test)

print("准确率:", accuracy_score(y_test, y_pred))
print(classification_report(y_test, y_pred))

准确率: 1.0

类别 Precision Recall F1-Score Support
0 1.00 1.00 1.00 2
1 1.00 1.00 1.00 2
accuracy 1.00 4
macro avg 1.00 1.00 1.00 4
weighted avg 1.00 1.00 1.00 4 
1
2
3
feature_importances = pd.Series(rf.feature_importances_, index=X.columns)
feature_importances = feature_importances.sort_values(ascending=False)
print(feature_importances)
特征 重要性
开盘 0.3772
最高 0.2516
最低 0.1511
收盘 0.1213
涨跌幅 0.0444
振幅 0.0196
成交量 0.0173
换手率 0.0095
成交额 0.0081

决策树

1
2
3
4
5
6
7
8
from sklearn.tree import DecisionTreeClassifier
dtc = DecisionTreeClassifier(random_state=42)
dtc.fit(X_train, y_train)
y_pred = dtc.predict(X_test)

from sklearn.metrics import classification_report, confusion_matrix
print(confusion_matrix(y_test, y_pred))
print(classification_report(y_test, y_pred))
实际 \ 预测 0 1
0 2 0
1 1 1
类别 Precision Recall F1-Score Support
0 0.67 1.00 0.80 2
1 1.00 0.50 0.67 2
accuracy 0.75 4
macro avg 0.83 0.75 0.73 4
weighted avg 0.83 0.75 0.73 4 
1
2
3
4
5
6
from sklearn.model_selection import cross_val_score

scores = cross_val_score(dtc, X_train, y_train, cv=5, scoring='accuracy')

print("各折准确率：", scores)
print("平均准确率：%.4f (+/- %.4f)" % (scores.mean(), scores.std() * 2))

各折准确率： [1. 1. 0.66666667 0.66666667 1. ]
平均准确率：0.8667 (+/- 0.3266)

决策树结果分析

发现决策树效果并没有随机森林那么好 这是因为随机森林是一种集成算法 是由多个决策树集成的 其结果是由多个决策树共同打分的 因此随机森林在处高维数据时会比决策树要好

GBDT

1
2
3
4
5
6
7
8
from sklearn.ensemble import GradientBoostingClassifier

GBDT = GradientBoostingClassifier(random_state=42)
GBDT.fit(X_train, y_train)
y_pred = GBDT.predict(X_test)

print(confusion_matrix(y_test, y_pred))
print(classification_report(y_test, y_pred))
实际 \ 预测 0 1
0 2 0
1 1 1
类别 Precision Recall F1-Score Support
0 0.67 1.00 0.80 2
1 1.00 0.50 0.67 2
accuracy 0.75 4
macro avg 0.83 0.75 0.73 4
weighted avg 0.83 0.75 0.73 4 
1
2
3
4
5
6
from sklearn.model_selection import cross_val_score

scores = cross_val_score(GBDT, X_train, y_train, cv=5, scoring='accuracy')

print("各折准确率：", scores)
print("平均准确率：%.4f (+/- %.4f)" % (scores.mean(), scores.std() * 2))

各折准确率： [1. 1. 0.66666667 0.66666667 1. ]
平均准确率：0.8667 (+/- 0.3266)

GBDT结果分析

发现GBDT的结果没随机森林要好 这是因为GBDT对参数敏感 需要进行调参 使用网格优化

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
from sklearn.model_selection import GridSearchCV

param_grid = {
    'n_estimators': [100, 200, 500],
    'learning_rate': [0.1, 0.05, 0.01],
    'max_depth': [3, 5, 7],
    'min_samples_split': [2, 5, 10],
    'min_samples_leaf': [1, 2, 4]
}

grid = GridSearchCV(
    GradientBoostingClassifier(random_state=42),
    param_grid,
    cv=5,
    scoring='f1_macro',  # 也可以换成 'accuracy', 'roc_auc'
    n_jobs=-1
)

grid.fit(X_train, y_train)
print("Best Params:", grid.best_params_)
print("Best Score:", grid.best_score_)

Best Params: {‘learning_rate’: 0.1, ‘max_depth’: 3, ‘min_samples_leaf’: 1, ‘min_samples_split’: 2, ‘n_estimators’: 100}
Best Score: 0.8666666666666666
其结果要比之前的好了一些

朴素贝叶斯

1
2
3
4
5
6
7
8
9
10
11
12
13
14
from sklearn.naive_bayes import GaussianNB

nb = GaussianNB()
nb.fit(X_train, y_train)

y_pred_nb = nb.predict(X_test)
y_pred_proba = nb.predict_proba(X_test)[:, 1]  # 离职概率（正类概率）

print("混淆矩阵：")
print(confusion_matrix(y_test, y_pred_nb))
print("\n分类报告：")
print(classification_report(y_test, y_pred_nb))

print("\n前5个样本的离职概率：", y_pred_proba[:5])
1
2
3
4
5
6
from sklearn.model_selection import cross_val_score

scores = cross_val_score(nb, X_train, y_train, cv=5, scoring='accuracy')

print("各折准确率：", scores)
print("平均准确率：%.4f (+/- %.4f)" % (scores.mean(), scores.std() * 2))

贝叶斯结果分析

可以看到 贝叶斯的结果非常不好 这是因为 贝叶斯模型是假设特征之间是相互独立的 而在股票市场中 特征之间往往是存在相关性的 因此贝叶斯模型的效果是要比其他模型差的

我们还可以画出特征之间的热力图

1
2
3
4
5
import seaborn as sns
correlation_matrix = pd.concat([X_train, y_train], axis=1).corr()

sns.heatmap(correlation_matrix, annot=True, cmap='coolwarm')
plt.show()

可以看到 像开盘与最高 开盘与最低等之间呈现高相关性

LDA

1
2
3
4
5
6
7
8
9
10
11
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis

LDA = LinearDiscriminantAnalysis()
LDA.fit(X_train, y_train)
y_pred_lda = LDA.predict(X_test)
y_pred_proba_lda = LDA.predict_proba(X_test)[:, 1]

print("混淆矩阵：")
print(confusion_matrix(y_test, y_pred_lda))
print("\n分类报告：")
print(classification_report(y_test, y_pred_lda))
实际 \ 预测 0 1
0 2 0
1 0 2
类别 Precision Recall F1-Score Support
0 1.00 1.00 1.00 2
1 1.00 1.00 1.00 2
accuracy 1.00 4
macro avg 1.00 1.00 1.00 4
weighted avg 1.00 1.00 1.00 4 
1
2
3
4
5
6
from sklearn.model_selection import cross_val_score

scores = cross_val_score(LDA, X_train, y_train, cv=5, scoring='accuracy')

print("各折准确率：", scores)
print("平均准确率：%.4f (+/- %.4f)" % (scores.mean(), scores.std() * 2))

各折准确率： [0.66666667 0.66666667 0.66666667 0.66666667 0.33333333]
平均准确率：0.6000 (+/- 0.2667)

LDA结果分析

我们看到 LDA的分类报告非常好 但是进行k折交叉验证的时候没那么好
这是因为我的数据样本少 分类报告恰好分类得很好 但是在交叉验证的时候发现了此问题 此外 对于股票行情 不推荐使用k折交叉验证 因为交叉验证是随机划分的 有可能会泄漏未来的信息 推荐使用时序交叉验证

1
2
3
4
5
6
from sklearn.model_selection import TimeSeriesSplit, cross_val_score

tscv = TimeSeriesSplit(n_splits=5)
scores = cross_val_score(LDA, X_train, y_train, cv=tscv, scoring='accuracy')
print("各折准确率：", scores)
print("平均准确率：%.4f (+/- %.4f)" % (scores.mean(), scores.std() * 2))

各折准确率： [0.5 0.5 0.5 0.5 0.5]
平均准确率：0.5000 (+/- 0.0000)

LDA 假设特征服从高斯分布 且协方差矩阵相同 但股票特征往往不满足

  • Title: 商业数据分析--股票涨跌预测模型搭建
  • Author: 姜智浩
  • Created at : 2025-09-25 11:45:14
  • Updated at : 2025-09-25 09:44:49
  • Link: https://super-213.github.io/zhihaojiang.github.io/2025/09/25/20250925商业数据分析--股票涨跌预测模型搭建/
  • License: This work is licensed under CC BY-NC-SA 4.0.
On this page
商业数据分析--股票涨跌预测模型搭建
  1. 声明
  2. 导入库
  3. 查看数据
  4. 画k线图
  5. 随机森林
  6. 决策树
    1. 决策树结果分析
  7. GBDT
    1. GBDT结果分析
  8. 朴素贝叶斯
    1. 贝叶斯结果分析
  9. LDA
    1. LDA结果分析