# Preprocessing question

**URL:** https://community.drivendata.org/t/preprocessing-question/2999
**Category:** Sustainable Industry: Rinse Over Run
**Created:** [January 12, 2019, 12:15am UTC](https://community.drivendata.org/t/preprocessing-question/2999 "2019-01-12T00:15:02Z")
**Posts on this page:** 4
**Page:** 1

<div class="post-metadata">

### Author: ![geomcintire](https://avatars.discourse-cdn.com/v4/letter/g/ac8455/32.png) [@geomcintire](https://community.drivendata.org/u/geomcintire)
#### Post date: [January 12, 2019, 12:15am UTC](https://community.drivendata.org/t/preprocessing-question/2999/1 "2019-01-12T00:15:02Z")

</div>

Here’s how I extracted the relevant data from “train\_values.csv”

Can someone confirm this is what we’re supposed to do in terms of readying the data for machine learning?  
Thanks. ![29%20PM](https://canada1.discourse-cdn.com/flex028/uploads/drivendata1/original/1X/f09915bd4eae15288fbd7d1e7fa7faa48e901b22.png)

---

<div class="post-metadata">

### Author: ![bull](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/bull/32/8_2.png) [@bull](https://community.drivendata.org/u/bull)
#### Post date: [January 14, 2019, 5:13pm UTC](https://community.drivendata.org/t/preprocessing-question/2999/2 "2019-01-14T17:13:45Z")

</div>

The benchmark notebook has an example of how to get started:  
[http://drivendata.co/blog/rinse-over-run-benchmark/](http://drivendata.co/blog/rinse-over-run-benchmark/)

---

<div class="post-metadata">

### Author: ![dcart](https://avatars.discourse-cdn.com/v4/letter/d/9e8a1a/32.png) [@dcart](https://community.drivendata.org/u/dcart)
#### Post date: [January 15, 2019, 7:44am UTC](https://community.drivendata.org/t/preprocessing-question/2999/3 "2019-01-15T07:44:44Z")

</div>

This is my first time to compete in time series forecasting.

I don’t expect someone to response, however, is this a multi-step forecasting problem?

---

<div class="post-metadata">

### Author: ![bull](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/bull/32/8_2.png) [@bull](https://community.drivendata.org/u/bull)
#### Post date: [January 16, 2019, 10:38pm UTC](https://community.drivendata.org/t/preprocessing-question/2999/4 "2019-01-16T22:38:11Z")

</div>

This is not a timeseries forecasting competition. Instead it is a regression question (what is the quantity of turbidity?) where the features that are used by the models are generated by timeseries. That said, you could do forecasting as an intermediate step but it is not strictly necessary. The linked benchmark will show you an example.