You wrote a pandas script to compute a fee on 50 million credit card transactions: 3% for foreign transactions, 1% otherwise, using a row-wise apply. On a 16GB laptop it ran for two hours and then crashed with an out-of-memory error. How would you diagnose and fix this?

你写了个pandas脚本，给5000万笔信用卡交易逐笔算手续费：境外交易收3%，其他收1%，用的是逐行apply。笔记本16G内存，跑了两个小时，最后内存溢出。你会怎么排查和优化？