给一个完整的决策树预测3D和值大小的源代码，并加入决策树规则

本帖最后由 杨哥哥儿 于 2025-6-18 21:26 编辑

code:

---

# -*- coding: utf-8 -*-
'''
完整的决策树预测3D彩票涨跌的源代码，并加入决策树规则。
'''
import pandas as pd
import numpy as np
from sklearn.tree import DecisionTreeClassifier, export_text
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
import matplotlib.pyplot as plt

# 获取彩票历史数据
def get_fc3d_data():
    # 读取文本文件（开奖文件），分隔符为空格，第一列为索引
    df = pd.read_csv(r'3d_asc.txt',delimiter=' ',index_col=0)
    # 删除不需要的数据，按列删除
    df.drop(df.columns[[4,5,6,7,8,9,10,11,12,13,14,15]], axis=1, inplace=True)#删除多余列
    # 设置各列的名称
    df.columns = ['日期', '百', '十','个']
    # 求和值
    df['和值']=df['百']+df['十']+df['个']    
    # 遍历每一行，处理并新增列
    for index, row in df.iterrows():
        df.loc[index, '跨度'] =np.abs(row['百'] - row['十'])+np.abs(row['百'] - row['个'])+np.abs(row['个'] - row['十'])
    return df

# 特征工程
def create_features(df):
    # 基础和值特征
    df['pct_change'] = df['和值'].pct_change()
    df['ma5'] = df['和值'].rolling(5).mean()
    df['ma10'] = df['和值'].rolling(10).mean()
    
    # 动量指标
    df['rsi'] = calculate_rsi(df['和值'])
    
    # 跨度关系
    
    df['vol_ma5'] = df['跨度'].rolling(5).mean()
    df['kua_he_corr'] = df['和值'].rolling(10).corr(df['跨度'])
    
    # 标记大小标签(1=大, 0=小)
    df['label'] = np.where(df['pct_change'].shift(-1) > 0, 1, 0)
    
    return df.dropna()

# 计算RSI指标
def calculate_rsi(prices, window=14):
    delta = prices.diff()
    gain = delta.where(delta > 0, 0)
    loss = -delta.where(delta < 0, 0)
    avg_gain = gain.rolling(window).mean()
    avg_loss = loss.rolling(window).mean()
    rs = avg_gain / avg_loss
    return 100 - (100 / (1 + rs))

# 决策树规则可视化
def show_rules(tree_model, feature_names):
    tree_rules = export_text(
        tree_model, 
        feature_names=list(feature_names),
        max_depth=3
    )
    print("决策树核心规则：\n", tree_rules)

if __name__ == '__main__':
    # 数据准备
    stock_data = get_fc3d_data()
    df = create_features(stock_data)
    
    # 特征选择
    features = ['ma5', 'ma10', 'rsi', 'vol_ma5', 'kua_he_corr']
    X = df[features]
    y = df['label']
    
    # 划分数据集
    X_train, X_test, y_train, y_test = train_test_split(
        X, y, test_size=0.2, random_state=42
    )
    
    # 构建决策树模型
    dt_model = DecisionTreeClassifier(
        max_depth=5,
        min_samples_split=20,
        criterion='gini',
        random_state=42
    )
    dt_model.fit(X_train, y_train)
    
    # 模型评估
    y_pred = dt_model.predict(X_test)
    print("测试集准确率:", accuracy_score(y_test, y_pred))
    
    # 输出决策规则
    show_rules(dt_model, features)
    
    # 特征重要性可视化
    plt.figure(figsize=(10,5))
    pd.Series(dt_model.feature_importances_, index=features)\
        .sort_values().plot(kind='barh')
    plt.title('Feature Importance')
    plt.show()

---

该代码实现了完整的股票预测流程，包含数据获取、特征计算和决策树建模。核心决策规则通过export_text函数可视化输出，特征重要性通过柱状图展示。
典型决策规则可能包含"RSI<30且跨度增加则看涨"等逻辑。
代码说明：
使用pd.read_csv(r'3d_asc.txt')数据，计算5/10日均线、RSI等技术指标作为特征，决策树通过基尼系数选择最优分割点，最终输出可解释的和值大小的开出规则。
你就可以根据AI算出的规则去预测下和值的开出情况了。
规则结果如下：

code:

---

测试集准确率: 0.6955997588908981
决策树核心规则：
 |--- rsi <= 51.58
|   |--- rsi <= 47.16
|   |   |--- rsi <= 44.22
|   |   |   |--- ma5 <= 14.50
|   |   |   |   |--- truncated branch of depth 2
|   |   |   |--- ma5 >  14.50
|   |   |   |   |--- truncated branch of depth 2
|   |   |--- rsi >  44.22
|   |   |   |--- ma10 <= 16.55
|   |   |   |   |--- truncated branch of depth 2
|   |   |   |--- ma10 >  16.55
|   |   |   |   |--- class: 0
|   |--- rsi >  47.16
|   |   |--- ma5 <= 13.30
|   |   |   |--- rsi <= 49.08
|   |   |   |   |--- truncated branch of depth 2
|   |   |   |--- rsi >  49.08
|   |   |   |   |--- truncated branch of depth 2
|   |   |--- ma5 >  13.30
|   |   |   |--- ma5 <= 16.10
|   |   |   |   |--- truncated branch of depth 2
|   |   |   |--- ma5 >  16.10
|   |   |   |   |--- truncated branch of depth 2
|--- rsi >  51.58
|   |--- ma5 <= 12.50
|   |   |--- rsi <= 53.42
|   |   |   |--- ma5 <= 11.10
|   |   |   |   |--- truncated branch of depth 2
|   |   |   |--- ma5 >  11.10
|   |   |   |   |--- truncated branch of depth 2
|   |   |--- rsi >  53.42
|   |   |   |--- ma10 <= 14.45
|   |   |   |   |--- truncated branch of depth 2
|   |   |   |--- ma10 >  14.45
|   |   |   |   |--- truncated branch of depth 2
|   |--- ma5 >  12.50
|   |   |--- rsi <= 53.46
|   |   |   |--- ma5 <= 16.10
|   |   |   |   |--- truncated branch of depth 2
|   |   |   |--- ma5 >  16.10
|   |   |   |   |--- truncated branch of depth 2
|   |   |--- rsi >  53.46
|   |   |   |--- rsi <= 58.64
|   |   |   |   |--- truncated branch of depth 2
|   |   |   |--- rsi >  58.64
|   |   |   |   |--- truncated branch of depth 2

---

算法是借股票的算法来完成的，可以较好的帮助大家理解。

更新了跨度计算公式，=|百-十|+|百-个|+|十-个|，没有了复杂的判断语句，值是跨度的2倍，不影响使用。

AI辅助彩票研究专题（六）常用机器学习算法与彩票的关联
这个专题明天再发帖。

就是看不懂