# Run model without Predicted field input

**URL:** <https://discourse.numenta.org/t/run-model-without-predicted-field-input/1930>\
**Category:** NuPIC\
**Tags:** question\
**Created:** [February 20, 2017, 5:33pm UTC](https://discourse.numenta.org/t/run-model-without-predicted-field-input/1930 "2017-02-20T17:33:55Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![Zizou](https://avatars.discourse-cdn.com/v4/letter/z/53a042/32.png) [@Zizou](https://discourse.numenta.org/u/Zizou)\
**Post date:** [February 20, 2017, 5:33pm UTC](https://discourse.numenta.org/t/run-model-without-predicted-field-input/1930/1 "2017-02-20T17:33:55Z")

</div>

Hi ,  
Firstly just like to say really like the HTM School tutorials !  
Just having first look at Nupic, I followed the One Hot Gym tutorial to test a different data set. The data is not time dependent and no time field, and swarm identified two fields that influence data which were two expected. Running the model and plotting the actual vs prediction looks to fit pretty well predicting one time step in advance. However I was interested to see how the predicted field would behave if seeding itself with its previous prediction (string value for binary classifier True False), the result was that it locked into the first value seeded and never changed. My question is is it possible to create a model that can freely predict a field only based on other influencing input values being fed for the hole sequence. For example share price predictor based on number of trades but never seeing any actual price (maybe not best example) Or is this against basic principles ?

---

<div class="post-metadata">

**Author:** ![rhyolight](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.numenta.org/rhyolight/32/3922_2.png) [@rhyolight](https://discourse.numenta.org/u/rhyolight)\
**Post date:** [February 20, 2017, 6:32pm UTC](https://discourse.numenta.org/t/run-model-without-predicted-field-input/1930/2 "2017-02-20T18:32:44Z")

</div>

> [@Zizou](#):
>
> My question is is it possible to create a model that can freely predict a field only based on other influencing input values being fed for the hole sequence. For example share price predictor based on number of trades but never seeing any actual price (maybe not best example) Or is this against basic principles ?

Yes you can do that, but you still need to encode the predicted field along with the other contributing fields. Read through this topic, and also [search the forums for more posts](https://discourse.numenta.org/search?q=multiple%20fields).

> [@How Should I Represent My Input with Multiple fields?](https://discourse.numenta.org/t/how-should-i-represent-my-input-with-multiple-fields/1234):
>
> I have some data that’s basically the vehicle flow over a bunch of road lanes in a 5 minute period with a timestamp (eg. lane 1: 10 cars, lane 2: 5, lane 3: 15 etc.) How should I best represent this as in input to my model using the OPF? My aim is to detect anomalies over the whole input. At first I thought I could just sum them all, but that gave poor results (see my last post), plus it’s intuitive to think that congestion in one lane will affect the flow in another lane (eg. a crash in a lane…

---

<div class="post-metadata">

**Author:** ![Zizou](https://avatars.discourse-cdn.com/v4/letter/z/53a042/32.png) [@Zizou](https://discourse.numenta.org/u/Zizou)\
**Post date:** [February 24, 2017, 3:01pm UTC](https://discourse.numenta.org/t/run-model-without-predicted-field-input/1930/3 "2017-02-24T15:01:57Z")

</div>

Thanks for feedback will take a look the multiple fields examples. Just to confirm , to encode the predicted field, should I update the model\_params.py , created by the the swam, for the predicted field in the sensorParams / Encoders section or should I be looking at creating dedicated encoder ?

---

<div class="post-metadata">

**Author:** ![rhyolight](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.numenta.org/rhyolight/32/3922_2.png) [@rhyolight](https://discourse.numenta.org/u/rhyolight)\
**Post date:** [February 24, 2017, 4:55pm UTC](https://discourse.numenta.org/t/run-model-without-predicted-field-input/1930/4 "2017-02-24T16:55:08Z")

</div>

If your predicted field is a scalar, you can just use a scalar encoder. No need to create an encoder unless you know how to extract domain-specific semantic information from your streams.

Yes, to include the predicted field it needs an encoder, but usually the swarm will encode the predicted field because its values contribute most to predictions.