sklearn 回归模型评价指标及代码示例

评价指标主要包括：

1. 回归评价指标

• R-Square （R^2）决定系数、线性相关系数
  
  取值范围为 [0,1] ，越接近1，模型效果越好，越接近0，效果越差；但是随着预测数据量的增加会增加降低R2值，因此只能大致评估模型

  from sklearn.metrics import r2_score
  print(r2_score(y_test,result_prediction))
  

• Mean Squared Error（MSE） 均方误差
  

  from sklearn.metrics import mean_squared_error
  print(mean_squared_error(y_test,result_prediction))
  

• Root Mean Squard Error（RMSE） 均方根误差
  

  import numpy as np
  from sklearn.metrics import mean_squared_error
  print(np.sqrt(mean_squared_error(y_test,result_prediction)))
  

• Mean Absolute Error（MAE） 平均绝对误差
  
  相当于求预测值与真实值的差值的绝对值，累加后求平均

  from sklearn.metrics import mean_absolute_error
  print(mean_absolute_error(y_test,result_prediction))
  

• Mean Absolute Percentage Error (MAPE)
  
  指标值域是[0,正无穷)，特别的，当数据中的“0”值较多时，这个指标会特别大，因此可以替换为SMAPE

  from sklearn.metrics import mean_absolute_percentage_error
  mape = mean_absolute_percentage_error(y_test, result_prediction)
  

• symmetric Mean Absolute Percentage Error (sMAPE)
  

  import numpy as np
  smape = 100 / len(y_test) * np.sum(
          2 * np.abs(result_prediction - y_test) / (np.abs(y_test) + np.abs(result_prediction)))
  

2. 整合工具函数

def regression_evaluation(true_value: np.ndarray, pred_value: np.ndarray) -> pd.DataFrame:
    """回归问题得到预测结果的指标
    :param true_value:真实值
    :param pred_value:预测值
    """
    from sklearn.metrics import mean_squared_error
    from sklearn.metrics import mean_absolute_error
    from sklearn.metrics import r2_score
    r2 = r2_score(true_value, pred_value)
    mae = mean_absolute_error(true_value, pred_value)
    rmse = np.sqrt(mean_squared_error(true_value, pred_value))
    mse = mean_squared_error(true_value, pred_value)
    smape = 100 / len(true_value) * np.sum(
        2 * np.abs(pred_value - true_value) / (np.abs(true_value) + np.abs(pred_value)))
    # MAPE指标
    # from sklearn.metrics import mean_absolute_percentage_error
    # mape = mean_absolute_percentage_error(true_value, pred_value)

    dataframe = pd.DataFrame([r2, mae, rmse, mse, smape]).T
    dataframe.columns = ['r2', 'MAE', 'RMSE', 'MSE', "SMAPE"]
    return dataframe

3. 全部指标&示例代码

import pandas as pd
import numpy as np


def regression_evaluation(true_value: np.ndarray, pred_value: np.ndarray) -> pd.DataFrame:
    """回归问题得到预测结果的指标
    :param true_value:真实值
    :param pred_value:预测值
    """
    from sklearn.metrics import mean_squared_error
    from sklearn.metrics import mean_absolute_error
    from sklearn.metrics import r2_score
    r2 = r2_score(true_value, pred_value)
    mae = mean_absolute_error(true_value, pred_value)
    rmse = np.sqrt(mean_squared_error(true_value, pred_value))
    mse = mean_squared_error(true_value, pred_value)
    smape = 100 / len(true_value) * np.sum(
        2 * np.abs(pred_value - true_value) / (np.abs(true_value) + np.abs(pred_value)))
    # MAPE指标
    # from sklearn.metrics import mean_absolute_percentage_error
    # mape = mean_absolute_percentage_error(true_value, pred_value)

    dataframe = pd.DataFrame([r2, mae, rmse, mse, smape]).T
    dataframe.columns = ['r2', 'MAE', 'RMSE', 'MSE', "SMAPE"]
    return dataframe


def get_data():
    """用来生成数据
    :return 
        ture_value:真实值
        pred:预测值
    """
    true_value = np.random.random(size=1000) * np.random.choice([-100, 100], size=1000)
    pred = true_value + np.random.random(size=1000) * np.random.choice([-1, 1], size=1000)
    return true_value, pred


def main():
    true_value, predict_value = get_data()
    re = regression_evaluation(true_value, predict_value)
    print(re)


if __name__ == '__main__':
    main()