使用Maximal Information Coefficient分析波士顿房价数据集遇minepy维度错误

Hey there, let's sort out this error you're hitting with minepy's MINE!

Why the Error Happens #

The compute_score() method from minepy.MINE expects two 1-dimensional arrays (one feature, one target) as inputs. But in your code, you're passing df[col_name[0:13]]—which is a 13-column DataFrame (a 2D structure)—as the first argument. That's exactly why you're getting the "Buffer has wrong number of dimensions" message.

Corrected Code #

Instead of passing all features at once, we need to loop through each feature individually and calculate its MIC with the target MEDV. Here's the fixed version:

import numpy as np
import pandas as pd
from minepy import MINE

# 读取数据集
df = pd.read_csv('housing.data', delim_whitespace=True, header=None)
col_name = ['CRIM', 'ZN' , 'INDUS', 'CHAS', 'NOX', 'RM', 'AGE', 'DIS', 'RAD', 'TAX', 'PTRATIO', 'B', 'LSTAT', 'MEDV']
df.columns = col_name

# 初始化MINE对象
m = MINE()

# 循环计算每个特征与MEDV的MIC
mic_results = {}
for feature in col_name[:-1]:  # 排除最后一个目标列MEDV
    m.compute_score(df[feature], df['MEDV'])
    mic_results[feature] = m.mic()

# 把结果转换成DataFrame，方便查看
mic_df = pd.DataFrame.from_dict(mic_results, orient='index', columns=['MIC'])
print(mic_df.sort_values(by='MIC', ascending=False))

What This Does #

• We iterate over each feature (excluding MEDV) and pass each as a 1D Series to compute_score() alongside the target MEDV.
• We We store each MIC value in a dictionary, then convert it to a sorted DataFrame so you can easily see which features have the highest mutual information with housing prices.

Bonus Tip #

If you want to avoid reinitializing the MINE object each time (though it's not strictly necessary), you can just reuse the same m instance as we did here—it works perfectly fine for multiple calculations.

内容的提问来源于stack exchange，提问作者Anx8