我构建了一个基于多个特征来预测房价的模型。

import statsmodels.api as statsmdlfrom sklearn import datasetsX = data[['NumberofRooms', 'YearBuilt','Type','NewConstruction']y = data["Price"]model = statsmdl.OLS(y, X).fit()predictions = model.predict(X)model.summary()

我如何找出这些特征中哪些是共线的？

回答：

您可以使用 DataFrame.corr() 方法。

示例：

In [27]: df = pd.DataFrame(np.random.randint(10, size=(5,3)), columns=list('abc'))In [28]: df['d'] = df['a'] * 10 - df['b'] / np.piIn [29]: df['e'] = np.log(df['c'] **2)In [30]: c = df.corr()In [31]: cOut[31]:          a         b         c         d         ea  1.000000  0.734858  0.113787  0.999837  0.067358b  0.734858  1.000000 -0.523635  0.722485 -0.598739c  0.113787 -0.523635  1.000000  0.129945  0.984257d  0.999837  0.722485  0.129945  1.000000  0.084615e  0.067358 -0.598739  0.984257  0.084615  1.000000In [32]: c[c >= 0.7]Out[32]:          a         b         c         d         ea  1.000000  0.734858       NaN  0.999837       NaNb  0.734858  1.000000       NaN  0.722485       NaNc       NaN       NaN  1.000000       NaN  0.984257d  0.999837  0.722485       NaN  1.000000       NaNe       NaN       NaN  0.984257       NaN  1.000000In [33]: c[c >= 0.7].stack().reset_index(name='cor').query("abs(cor) < 1.0")Out[33]:   level_0 level_1       cor1        a       b  0.7348582        a       d  0.9998373        b       a  0.7348585        b       d  0.7224857        c       e  0.9842578        d       a  0.9998379        d       b  0.72248511       e       c  0.984257