Merge pull request '预测企业所得税' (#1) from 预测企业所得税 into main

8.2.2 计算Pearson相关系数.py

@@ -0,0 +1,44 @@
+import numpy as np
+import pandas as pd
+from pyecharts import options as opts
+from pyecharts.charts import HeatMap
+
+# 修改为正确的路径
+income_data = pd.read_csv('data/income_tax.csv', index_col=0)  # 使用相对路径
+
+# 保留两位小数
+data_cor = np.round(income_data.corr(method = 'pearson'), 2)
+
+y_data = list(data_cor.columns)  # 获取y轴标签
+x_data = list(data_cor.index)  # 获取x轴标签
+# 相关系数矩阵转为列表
+values = data_cor.values.tolist()
+# 对应相关系数的位置
+value = [[i, j, values[i][j]] for i in range(len(x_data))
+         for j in range(len(y_data))]
+
+heatmap = (
+    # 导入热力图
+    HeatMap()
+    # 设置x轴
+    .add_xaxis(x_data)
+    # 设置y轴
+    .add_yaxis(
+        '', y_data, 
+        value, 
+        label_opts=opts.LabelOpts(
+            is_show=True, position='inside'),
+    )
+    .set_global_opts(
+        # 设置标题
+        title_opts=opts.TitleOpts(title='相关系数热力图'),
+        # 设置图例
+        visualmap_opts=opts.VisualMapOpts(
+            is_show=False, pos_bottom='center',
+            max_=1, min_=0.9
+        )
+    )
+)  
+
+# 保存为HTML文件的路径
+heatmap.render('tmp/相关系数热力图.html')  # 使用相对路径保存文件

8.3.2 选取关键特征.py

@@ -0,0 +1,35 @@
+import pandas as pd
+import numpy as np
+from sklearn.linear_model import Lasso
+
+# 修改为正确的路径
+income_data = pd.read_csv('data/income_tax.csv', index_col=0)  # 读取数据
+
+# 取2005年~2019年的数据建模
+data_train = income_data.iloc[:, 0:13].copy()  
+data_mean = data_train.mean()
+data_std = data_train.std()
+data_train = (data_train - data_mean) / data_std  # 数据标准化
+
+# 构建Lasso回归模型
+lasso = Lasso(alpha=1000, random_state=1234)
+lasso.fit(data_train, income_data['y'])
+
+# 输出回归系数，保留5位小数
+print('Lasso回归系数为：', np.round(lasso.coef_, 5))
+
+# 计算系数非零的个数
+print('系数非零个数为：', np.sum(lasso.coef_ != 0))  
+
+# 返回系数非零特征
+mask = lasso.coef_ != 0
+print('系数非零特征：', income_data.columns[:-1][mask])
+
+# 返回系数非零的数据
+new_reg_data = income_data.iloc[:, 0:13].iloc[:, mask]
+
+# 修改为正确的路径保存数据
+new_reg_data.to_csv('tmp/new_reg_data.csv')  # 存储数据
+
+# 输出数据的维度
+print('输出数据的维度为：', new_reg_data.shape)

8.4.3 预测企业所得税.py

@@ -0,0 +1,94 @@
+import numpy as np
+import pandas as pd
+from gm11 import gm11  # 引入自编的灰色预测函数
+
+# 读取经过特征选择后的数据
+new_reg_data = pd.read_csv('tmp/new_reg_data.csv', index_col=0)  # 修改为相对路径
+# 读取数据
+income_data = pd.read_csv('data/income_tax.csv', index_col=0)  # 修改为相对路径
+new_reg_data.index = range(2005, 2020)
+new_reg_data.loc[2020] = None
+new_reg_data.loc[2021] = None
+
+c = []
+p = []
+# 进行灰色预测
+for i in list(new_reg_data.columns):
+    f = gm11(np.array(new_reg_data.loc[range(2005, 2020), i]))[0]
+    c.append(gm11(np.array(new_reg_data.loc[range(2005, 2020), i]))[4])
+    p.append(gm11(np.array(new_reg_data.loc[range(2005, 2020), i]))[5])
+    new_reg_data.loc[2020, i] = f(len(new_reg_data) - 1)
+    new_reg_data.loc[2021, i] = f(len(new_reg_data))
+    new_reg_data[i] = new_reg_data[i].round(2)  # 保留两位小数
+
+new_reg_data = pd.concat([new_reg_data, income_data['y']], axis=1)
+new_reg_data.to_csv('tmp/new_reg_data_GM11.csv')  # 保存为相对路径
+print('预测结果为：\n', new_reg_data.iloc[-2:, :6])  # 预测结果展示
+
+from sklearn.svm import LinearSVR
+from sklearn.metrics import explained_variance_score, mean_absolute_error, median_absolute_error, r2_score
+
+# 读取灰色预测数据
+gm11_data = pd.read_csv('tmp/new_reg_data_GM11.csv', index_col=0)  # 修改为相对路径
+feature = gm11_data.columns[:-1]  # 获取特征列（排除最后一列'y'）
+
+# 取2005~2019年的数据建模
+data_train = gm11_data.loc[range(2005, 2020)].copy()
+data_mean = data_train.mean()
+data_std = data_train.std()
+data_train = (data_train - data_mean) / data_std  # 数据标准化
+
+x_train = np.array(data_train[feature])  # 特征数据
+y_train = np.array(data_train['y'])  # 标签数据
+linearsvr = LinearSVR(random_state=1234)  # 调用LinearSVR类
+linearsvr.fit(x_train, y_train)
+
+# 预测，并还原结果
+x = np.array(((gm11_data[feature] - data_mean[feature]) / data_std[feature]))
+gm11_data['y_pred'] = linearsvr.predict(x) * data_std['y'] + data_mean['y']
+
+# 保存预测结果
+gm11_data.to_csv('tmp/new_reg_data_GM11_revenue.csv')  # 保存为相对路径
+print('真实值与预测值分别为：\n', gm11_data[['y', 'y_pred']])
+
+# 计算各项评估指标
+print('可解释方差值：', explained_variance_score(gm11_data['y'][:-2], gm11_data['y_pred'][:-2]))
+print('平均绝对误差：', mean_absolute_error(gm11_data['y'][:-2], gm11_data['y_pred'][:-2]))
+print('中值绝对误差：', median_absolute_error(gm11_data['y'][:-2], gm11_data['y_pred'][:-2]))
+print('R2值：', r2_score(gm11_data['y'][:-2], gm11_data['y_pred'][:-2]))
+
+from pyecharts.charts import Grid, Scatter, Line
+from pyecharts import options as opts
+
+# 设置x轴的值
+x_data = ['2005', '2006', '2007', '2008', '2009', '2010',
+          '2011', '2012', '2013', '2014', '2015', '2016',
+          '2017', '2018', '2019', '2020', '2021']
+
+# 绘制线
+line = (Line(init_opts=opts.InitOpts(width='800px', height='310px'))
+   .add_xaxis(x_data)
+   .add_yaxis('真实值', gm11_data['y'].tolist(), label_opts=opts.LabelOpts(is_show=False)) 
+   .add_yaxis('预测值', gm11_data['y_pred'].tolist(), label_opts=opts.LabelOpts(is_show=False))
+)
+
+# 绘制点
+scatter = (
+    Scatter(init_opts=opts.InitOpts(width='800px', height='310px'))
+    .add_xaxis(x_data)
+    .add_yaxis('真实值', gm11_data['y'].tolist(), label_opts=opts.LabelOpts(is_show=False), symbol_size=10, symbol='diamond')
+    .add_yaxis('预测值', gm11_data['y_pred'].tolist(), label_opts=opts.LabelOpts(is_show=False), symbol_size=10, symbol='pin')
+    .set_global_opts(
+        title_opts=opts.TitleOpts(title='真实值与预测值对比'),
+        yaxis_opts=opts.AxisOpts(name='企业所得税（万元）', name_location='middle', name_gap=70),
+        xaxis_opts=opts.AxisOpts(name='年份', name_location='middle', name_gap=30),
+    )
+)
+
+# 叠加图
+scatter.overlap(line)
+
+# 创建网格并保存结果
+grid = Grid()
+grid.add(scatter, grid_opts=opts.GridOpts(pos_top='10%', pos_left='12%', pos_bottom='35%'))
+grid.render('tmp/真实值与预测值对比.html')  # 保存为相对路径