尽管已删除，但 Python 随机森林回归器仍对 nan 值出错

Question

我有一个干净的数据集，其 nan 值为零，但我继续在回归器上遇到相同的错误。我的框架叫做 new_player_data我试过找到任何list(new_player_data.where(new_player_data.isna()).count() > 0)返回[假，假，假，假，假，假]大约两百次。我认为可能有一些太大的浮动。我试过这个：for i in new_player_data.columns[:]:    if new_player_data[i].dtype == float:        new_player_data[i] = round(new_player_data[i],2)无论我得到什么：regressor.fit(X_train, y_train)  ---------------------------------------------------------------------------ValueError                                Traceback (most recent call last)<ipython-input-327-3a664017ddaa> in <module>----> 1 regressor.fit(X_train, y_train)/anaconda3/lib/python3.7/site-packages/sklearn/ensemble/forest.py in fit(self, X, y, sample_weight)    248     249         # Validate or convert input data--> 250         X = check_array(X, accept_sparse="csc", dtype=DTYPE)    251         y = check_array(y, accept_sparse='csc', ensure_2d=False, dtype=None)    252         if sample_weight is not None:/anaconda3/lib/python3.7/site-packages/sklearn/utils/validation.py in check_array(array, accept_sparse, accept_large_sparse, dtype, order, copy, force_all_finite, ensure_2d, allow_nd, ensure_min_samples, ensure_min_features, warn_on_dtype, estimator)    571         if force_all_finite:    572             _assert_all_finite(array,--> 573                                allow_nan=force_all_finite == 'allow-nan')    574     575     shape_repr = _shape_repr(array.shape)/anaconda3/lib/python3.7/site-packages/sklearn/utils/validation.py in _assert_all_finite(X, allow_nan)     54                 not allow_nan and not np.isfinite(X).all()):     55             type_err = 'infinity' if allow_nan else 'NaN, infinity'---> 56             raise ValueError(msg_err.format(type_err, X.dtype))     57      58 ValueError: Input contains NaN, infinity or a value too large for dtype('float32').关于我还可以在这里检查什么的任何想法？亏本

Answer 1

发现它是 inf 值，通过

infs = np.where(np.isinf(new_player_data))

infs

out: (array([ 261, 1162, 1190, 1339, 1365, 1451, 1656, 1736, 1878, 1954, 2189,

    2299, 2741, 3137, 3162, 3799, 3821, 3881, 4305]),

 array([ 3, 43, 43,  3, 43, 43, 43, 43, 43, 43, 23, 43,  3, 43, 43, 43,  3,

    23, 43]))

然后我就这样替换了它们

pd.options.mode.use_inf_as_na = True

infs = np.where(np.isinf(new_player_data))

infs

out: (array([], dtype=int64), array([], dtype=int64))