python 根据文本相似度：数据去重

1，需求

文本列表如下：

SQL	平均耗时	执行次数
SELECT DISTINCT object from t1	5000	23
SELECT DISTINCT object as cnt from t1	5132	12
SELECT COUNT(*) FROM t3	5678	56
SELECT COUNT(*) as cnt FROM t3	5001	3

经python 脚本处理：结果如下

SQL	平均耗时	执行次数
SELECT DISTINCT object from t1	5066	35
SELECT COUNT(*) FROM t3	5339.5	59

2，python 根据相似度去重，并归类统计

import pandas as pd  
from fuzzywuzzy import fuzz  
from collections import defaultdict  
  
def read_excel(file_path):  
    # 读取Excel文件  
    df = pd.read_excel(file_path)  
    return df  
  
def process_data(df):  
    # 使用defaultdict来存储处理后的数据  
    processed_data = defaultdict(lambda: {'sum_col3': 0, 'avg_col2': 0, 'count_col2': 0})  
      
    # 遍历DataFrame的每一行  
    for index, row in df.iterrows():  
        # 假设第一列是text_col，第二列是num_col2，第三列是num_col3  
        text_col = row[0]  
        #print(text_col)
        num_col2 = row[1]  
        num_col3 = row[2]  
          
        # 查找最相似的项  
        max_ratio = 0  
        best_match = None  
        for key in processed_data:  
            ratio = fuzz.ratio(text_col, key)  
            if ratio > max_ratio:  
                max_ratio = ratio  
                best_match = key  
          
        # 如果找到了足够相似的项（这里我们假设60%相似度是足够的）  
        if max_ratio > 60:  
            # 更新第二列的和及计数  
            processed_data[best_match]['sum_col2'] += num_col2  
            processed_data[best_match]['count_col2'] += 1  
            # 更新第三列的和  
            processed_data[best_match]['sum_col3'] += num_col3  
        else:  
            # 如果没有足够相似的项，则作为一个新项添加  
            processed_data[text_col]['sum_col2'] = num_col2  
            processed_data[text_col]['count_col2'] = 1  
            processed_data[text_col]['sum_col3'] = num_col3  
      
    # 计算第二列的平均值  
    for key, value in processed_data.items():  
        if value['count_col2'] > 0:  
            value['avg_col2'] = value['sum_col2'] / value['count_col2']  
      
    # 转换为DataFrame以便输出  
    result_df = pd.DataFrame.from_dict(processed_data, orient='index')  
    result_df.reset_index(inplace=True)  
    return result_df  
  
# 结果另存为csv文件  
file_path = '/home/test/SQL.xlsx'  # 替换为你的Excel文件路径  
df = read_excel(file_path)  
result_df = process_data(df)  
#print(result_df)j
# index   sum_col3       avg_col2  count_col2      sum_col2
df2=result_df[['index', 'avg_col2','sum_col3']]  
df2.to_csv('example.csv', index=True)  # index=False表示不将行索引写入文件