个人博客转至:tybai.com

无聊就想打码,打码使我快乐


Fork me on GitHub

pandas的apply操作

pandas的apply操作类似于Scala的udf一样方便,假设存在如下dataframe

  id_part                  pred               pred_class v_id
0       d  [0.722817, 0.650064]                  cat,dog   d1
1       5  [0.119208, 0.215449]  other_label,other_label   d2

需要把 v_id=d1 中,predpred_class 一一对应,需要将 pred 大于0.5的pred_class取出来作为新的一列,如果小于0.5则不取出来:

import pandas as pd


# 提取类别
def get_pred_class(pred_class, pred):
    pred_class_list = pred_class.split(",")
    result_class_list = []
    for i in range(0, len(pred)):
        if float(pred[i]) >= 0.5:
            result_class_list.append(pred_class_list[pred.index(pred[i])])
    return result_class_list


# 新建一个dataframe
data = pd.DataFrame({
    'v_id': ["d1", 'd2'],
    'pred_class': ["cat,dog", 'other_label,other_label'],
    'pred': [[0.722817,0.650064], [0.119208,0.215449]],
    'id_part': ["d", '5'],
})

df = data.copy()
df["pos_labels"] = data.apply(lambda row: get_pred_class(row['pred_class'], row['pred']), axis=1)
print(df)

得到结果为:

  id_part                  pred               pred_class v_id  pos_labels
0       d  [0.722817, 0.650064]                  cat,dog   d1  [cat, dog]
1       5  [0.119208, 0.215449]  other_label,other_label   d2          []

PS:如果没有df = data.copy()将会出现错误:

A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

posted on 2018-09-28 11:13 TTyb 阅读(...) 评论(...) 编辑 收藏

导航


不用多久

我就会升职加薪

当上总经理

出任CEO

迎娶白富美

走上人生巅峰

Pulpit rock