我正在嘗試使用諸如分位數范圍和 zscore 之類的統計方法創建一個類來查找資料集中的例外值。我想知道為什么有些例外值會從我的 pd.Series 串列中洗掉，而有些例外值沒有給出空條件。
class OutliersDetector:
def __init__(self, X):
self.outliers = []
self.X = X
def detect_range(self):
self.reset_outliers()
# to implement
self.remove_empty_items()
def detect_zscore(self):
self.reset_outliers()
zscore = np.abs(stats.zscore(self.X))
threshold_std = 3
for index, col_name in enumerate(self.X.columns): # X and zscore always have the same shape
col = zscore[:, index]
self.outliers.append( pd.Series(col[col >= threshold_std], name=col_name) )
self.remove_empty_items()
# none of the if statements i tried worked
def remove_empty_items(self):
for index, item in enumerate(self.outliers):
#if item.size == 0:
#if len(item.index) == 0:
if item.empty:
print("[no outliers] {}".format(item.name))
self.outliers.pop(index)
def reset_outliers(self):
self.outliers = []
def show_outliers(self):
for item in self.outliers:
print("[name]: {}\n[outliers]: {}\n".format(item.name, item.size))
outliers_detector = OutliersDetector(X_train_transformed)
outliers_detector.detect_zscore()
print("\noutliers found: ")
outliers_detector.show_outliers()
輸出：Rainfall、Month、Location、WindDir9a 不應列印在“發現例外值”下方，因為大小為 0，但...
[no outliers] RainToday
[no outliers] Year
[no outliers] Day
[no outliers] WindGustDir
[no outliers] WindDir3pm
[no outliers] Sunshine
[no outliers] Humidity3pm
[no outliers] Cloud9am
outliers found:
[name]: Rainfall
[outliers]: 0
[name]: Evaporation
[outliers]: 289
[name]: Month
[outliers]: 0
[name]: Location
[outliers]: 0
[name]: WindDir9am
[outliers]: 0
我怎樣才能解決這個問題？
uj5u.com熱心網友回復：
在remove_empty_items您self.outliers在迭代串列時修改串列。這會導致未定義的行為。您的代碼應該創建一個新串列，而不是修改當前串列：
def remove_empty_items(self):
non_empty_outliers = []
for item in self.outliers:
if item.empty:
print("[no outliers] {}".format(item.name))
else:
non_empty_outliers.append(item)
self.outliers = non_empty_outliers
轉載請註明出處，本文鏈接：https://www.uj5u.com/qukuanlian/432213.html
上一篇：如何使用全域值列印類中的屬性？