Python sklearn OneClassSVM只给出误报

我正在编写一个简单的API来输出一维数据集中的异常值。但出于某种原因，当数据使人认为它应该做相反的事情时，sklearn.svm.OneClassSVM只输出误报。下面是一个有问题的例子：

select valor, count(*) as no_occurrences
from medicao_eta
where dt_registro_programada between '2020-10-01 00:00:00:000' and '2020-10-13 23:59:59:997'
and cod_configuracao = 1331
group by valor 
order by count(*) desc

这给了我293条记录，只有两个值：

+--------+----------------+
| valor  | no_occurrences |
+--------+----------------+
| 1.3220 |            292 |
| 1.3200 |              1 |
+--------+----------------+

python代码：

def why_on_earth():
    import datetime
    import numpy as np
    from sklearn.svm import OneClassSVM

    db = DbUtil()

    unidade_operacional = db.fetch_unidade_operacional(73)
    test_data = db.fetch_medicoes(unidade_operacional=unidade_operacional,
        data_inicio=datetime.datetime(2020,10,1),
        data_fim=datetime.datetime(2020,10,13),
        cod_configuracao=1331)

    data = [row.valor for row in test_data[1331]]
    x_numpy = np.asarray(data)
    x = x_numpy.reshape(-1, 1)
    svm = OneClassSVM(kernel='rbf', gamma='scale', nu=0.001)
    pred = svm.fit_predict(x)

    indexes = [i for i, x in enumerate(pred) if x == -1]
    outliers = [test_data[1331][i] for i in indexes]

    return outliers

我已经检查了流入算法的数据一千次了。我看到293个小数组成了一个二维数组，比如

[[1.322]，[1.322]，[1.322]…]

。该算法不应该只输出一个异常值（1.3200）吗？pred数组只包含-1，这根本没有意义。我真的被困在这里了

以下是数据：

提前谢谢