如何给CSV文件（pandas读取的）的某列进行onehot编码

前言

在用mini-imagenet数据集时，要对csv文件的labels的列进行onehot编码，网上没怎么查到，顺手写了一波。

---

上代码

1.引入库

代码如下（示例）：

from numpy import argmax 
import pandas as pd
import numpy as np

2.onehot编码

代码如下（示例）：

new_train_path = r'D:\CsDataset\mini-imagenet\new_train.csv'

class OnehotLables:
    def __init__(self, labrel_idx):
        self.label_idx = labrel_idx   # 要编码的那列索引
        
    def _creat_one_hot(self,labels_len):   # labels_len ：要编码的列长
        '''做一个单位阵形式的列表'''
        return np.eye(labels_len).tolist()
    
    def _create_onehot_labels(self, labels, labels_set_list): #labels_set_list 取那列不重复的形成列表
        '''将要编码那列编码'''
        labels_len = len(labels_set_list)
        onehot = self._creat_one_hot(labels_len)
        for i in range(len(onehot)):
            for j in range(len(labels)):
                if labels[j] == labels_set_list[i]:
                    labels[j] = onehot[i]
        return labels
    
    def get_label_name(self, onehot_label, labels_set_list):
        '''取编码对应编码前的值'''
        return labels_set_list[argmax(onehot_label)]
    
    def forward(self, csv_path):  #  传入文件路径
        csv_data = pd.read_csv(csv_path)
        labels = csv_data.iloc[:, self.label_idx]
        labels_set_list = list(set(labels))
        # print(f'num_classification: {len(labels_set_list)}')
        onehot_labels = self._create_onehot_labels(labels, labels_set_list)
        return onehot_labels, labels_set_list

creat_onehot_labels = OnehotLables(labrel_idx=1)
onehot_labels, labels_set_list = creat_onehot_labels.forward(new_train_path)

onehot_label = onehot_labels[0]
print(onehot_label)
print(creat_onehot_labels.get_label_name(onehot_label,labels_set_list))

---

结果：

(base) C:\Users\渺渺夕\Desktop\Transfer_Learning_learning>D:/Anaconda3/python.exe c:/Users/渺渺夕/Desktop/Transfer_Learning_learning/data.py
[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
n01532829