# Evaluation Data ——Financial Crime

**URL:** <https://community.drivendata.org/t/evaluation-data-financial-crime/8286>\
**Category:** PETs Prize Challenge\
**Created:** [December 13, 2022, 7:45am UTC](https://community.drivendata.org/t/evaluation-data-financial-crime/8286 "2022-12-13T07:45:59Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![zhang66](https://avatars.discourse-cdn.com/v4/letter/z/c89c15/32.png) [@zhang66](https://community.drivendata.org/u/zhang66)\
**Post date:** [December 13, 2022, 7:45am UTC](https://community.drivendata.org/t/evaluation-data-financial-crime/8286/1 "2022-12-13T07:45:59Z")

</div>

We found the number of test data per day is half of train data:  
1)the evaluation data is the remaining data of the test data?  
2)feature enginnering can use now test data？

---

<div class="post-metadata">

**Author:** ![jayqi](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/jayqi/32/924_2.png) [@jayqi](https://community.drivendata.org/u/jayqi)\
**Post date:** [December 14, 2022, 3:30pm UTC](https://community.drivendata.org/t/evaluation-data-financial-crime/8286/2 "2022-12-14T15:30:53Z")

</div>

Hi @zhang66,

This is not a normal machine learning challenge where you train a model on training data and submit predictions on test data. In this challenge, teams submit code that will be executed to perform both training and test in a remote evaluation cluster on held out data. The evaluation data is an entirely separate and disjoint dataset consistent of both train and test splits.

Please review the documentation here: [Competition: U.S. PETs Prize Challenge: Phase 2 (Financial Crime)](https://www.drivendata.org/competitions/105/nist-federated-learning-2-financial-crime-federated/page/589/#development-and-evaluation-data)

---

<div class="post-metadata">

**Author:** ![kevinchow32](https://avatars.discourse-cdn.com/v4/letter/k/838e76/32.png) [@kevinchow32](https://community.drivendata.org/u/kevinchow32)\
**Post date:** [December 17, 2022, 1:11pm UTC](https://community.drivendata.org/t/evaluation-data-financial-crime/8286/3 "2022-12-17T13:11:07Z")

</div>

Hi @jayqi ,

During evaluation phase, is the test split accessible during training? If yes, what information is accessible?

Thanks,

---

<div class="post-metadata">

**Author:** ![jayqi](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/jayqi/32/924_2.png) [@jayqi](https://community.drivendata.org/u/jayqi)\
**Post date:** [December 17, 2022, 9:47pm UTC](https://community.drivendata.org/t/evaluation-data-financial-crime/8286/4 "2022-12-17T21:47:37Z")

</div>

Hi @kevinchow32,

During the evaluation, the test split will not be made accessible to the training code. Please see the documentation for how the data will be provided to your training code: [Competition: U.S. PETs Prize Challenge: Phase 2 (Financial Crime)](https://www.drivendata.org/competitions/105/nist-federated-learning-2-financial-crime-federated/page/587/)