匹配两个具有相似但不完全匹配数据行的数据库中的名称

Question

在此问题中，我尝试使用difflib和Fuzzy wuzzy来匹配名称，但是由于名称变化，匹配率很差。我现在正尝试将我拥有的其他数据字段用于名称，但是完全不确定如何处理此类问题。如果我不清楚，请告诉我，我会尽力澄清。

我有两个数据框，它们具有相似但不完全匹配的有关人的信息。我正在寻找使每个数据框的参考号与另一个数据框相匹配的参考号，这是每个人唯一的。例如，在下表中，我想知道第一个数据帧中Jimmy / James Random的引用号（因为他们是同一个人，但名字不匹配）在DF1中为1234，在DF2中为89。请注意，一个人的等级可能会发生变化，但是会同时在两个表中发生变化。每个人的参考编号，样式，ID和国籍将始终保持不变。

df1 = pd.DataFrame(columns=["Ref","Date","Name", "Rank","Nationality","Style","ID"], \
                      data=[["1234","20200104","Jimmy Random","General","France","Aggressive",""],\
                           ["1333","20200104","Ian Fleming","Brigadier","England","Passive","14"],\
                           ["1234","20191204","Jimmy Random","Major","France","","15"],\
                           ["1000","20200404","Peter Nisbett","Corporal","","Passive","12"]])
df2 = pd.DataFrame(columns=["Ref","Date","Name", "Rank","Nationality","Style","ID"], \
                      data=[["89","20200104","James Random","","France","Aggressive","104"],\
                           ["10","20200104","I. Fleming","Brigadier","England","","4"],\
                           ["156","20200404","P. Nisbett","","Spain","Passive","5"],\
                           ["89","20191204","James Random","Major","France","Aggressive","104"]])

非常感谢您提供的帮助。

芝士汉堡

Answer 1

您基本上需要将字符串与其他分析进行比较，对吗？检查余弦相似度，通过scit-learn实现。