# Guessing last column of data **URL:** \ **Category:** DL4J\ **Created:** [October 4, 2021, 11:29am UTC](https://community.konduit.ai/t/guessing-last-column-of-data/1628 "2021-10-04T11:29:25Z")\ **Posts on this page:** 8\ **Page:** 1
**Author:** ![mustafasari](https://yyz1.discourse-cdn.com/flex035/user_avatar/community.konduit.ai/mustafasari/32/645_2.png) [@mustafasari](https://community.konduit.ai/u/mustafasari)\ **Post date:** [October 4, 2021, 11:29am UTC](https://community.konduit.ai/t/guessing-last-column-of-data/1628/1 "2021-10-04T11:29:25Z")
I have a data like this : > My data columns are : from\_id,subject\_id,user\_id,user\_group\_id,sent\_user And i want to guess and suggest next sent\_user from his previous processes. So thats the problem. i want to guess to last column(sent\_user) but i couldnt find best way with d4lj, should i use Paragraph Vector? I did it with spark but i wonder what is the best way to guess t with dl4j. Any help is appreciated. ---
**Author:** ![treo](https://yyz1.discourse-cdn.com/flex035/user_avatar/community.konduit.ai/treo/32/47_2.png) [@treo](https://community.konduit.ai/u/treo)\ **Post date:** [October 4, 2021, 11:35am UTC](https://community.konduit.ai/t/guessing-last-column-of-data/1628/2 "2021-10-04T11:35:34Z")
> [@mustafasari](#): > > should i use Paragraph Vector What made you think that Paragraph Vectors of all things would be the reasonable approach here? There are a few possible approaches, but without knowing more about your data and the problem you are trying to solve it will be hard to point you at anything specific. Anyway, You can probably get a better idea how to approach things by reading my quickstart guide: [Quickstart with Deeplearning4J – dubs·tech](https://www.dubs.tech/guides/quickstart-with-dl4j/) ---
**Author:** ![mustafasari](https://yyz1.discourse-cdn.com/flex035/user_avatar/community.konduit.ai/mustafasari/32/645_2.png) [@mustafasari](https://community.konduit.ai/u/mustafasari)\ **Post date:** [October 4, 2021, 11:43am UTC](https://community.konduit.ai/t/guessing-last-column-of-data/1628/3 "2021-10-04T11:43:08Z")
My data columns are : from\_id,subject\_id,user\_id,user\_group\_id,sent\_user And i want to guess and suggest next sent\_user from his previous processes. So thats the problem. ---
**Author:** ![treo](https://yyz1.discourse-cdn.com/flex035/user_avatar/community.konduit.ai/treo/32/47_2.png) [@treo](https://community.konduit.ai/u/treo)\ **Post date:** [October 4, 2021, 11:52am UTC](https://community.konduit.ai/t/guessing-last-column-of-data/1628/4 "2021-10-04T11:52:12Z")
So all of your columns are categorical. You will want to convert them to one-hot encodings. And your output is also categorical, so you have a classification problem. However, I guess you want to be able to do the prediction with users you haven’t seen yet. My best guess is that what you are actually trying to build is a recommender system, which is a topic that goes way beyond simple regression or classification tasks. In that case DL4J would at best be just a small part of your solution. ---
**Author:** ![mustafasari](https://yyz1.discourse-cdn.com/flex035/user_avatar/community.konduit.ai/mustafasari/32/645_2.png) [@mustafasari](https://community.konduit.ai/u/mustafasari)\ **Post date:** [October 4, 2021, 11:59am UTC](https://community.konduit.ai/t/guessing-last-column-of-data/1628/5 "2021-10-04T11:59:32Z")
Like e-mail sent suggestions but its not only frequent sent users. It analyses whole data of all users, it include body , from,co-work or not, subject… at last it suggest user to you might sent tis mail to this users… because other users sent this subject and this body like mail to that users… ---
**Author:** ![treo](https://yyz1.discourse-cdn.com/flex035/user_avatar/community.konduit.ai/treo/32/47_2.png) [@treo](https://community.konduit.ai/u/treo)\ **Post date:** [October 4, 2021, 12:04pm UTC](https://community.konduit.ai/t/guessing-last-column-of-data/1628/6 "2021-10-04T12:04:57Z")
Yes, that is what you’d call a recommender system. We can help you with specific problems you run into when using DL4J here, but building an entire recommendation system to your specifications is far outside the scope here. You’ll need to do your own work. ---
**Author:** ![mustafasari](https://yyz1.discourse-cdn.com/flex035/user_avatar/community.konduit.ai/mustafasari/32/645_2.png) [@mustafasari](https://community.konduit.ai/u/mustafasari)\ **Post date:** [October 4, 2021, 1:23pm UTC](https://community.konduit.ai/t/guessing-last-column-of-data/1628/7 "2021-10-04T13:23:18Z")
No i just wanted a method or way… Or example on dl4j. btw some links are broken on the home page like [https://deeplearning4j.konduit.ai/getting-started/tutorials](https://deeplearning4j.konduit.ai/getting-started/tutorials) [https://deeplearning4j.konduit.ai/support](https://deeplearning4j.konduit.ai/support) ---
**Author:** ![treo](https://yyz1.discourse-cdn.com/flex035/user_avatar/community.konduit.ai/treo/32/47_2.png) [@treo](https://community.konduit.ai/u/treo)\ **Post date:** [October 4, 2021, 5:29pm UTC](https://community.konduit.ai/t/guessing-last-column-of-data/1628/8 "2021-10-04T17:29:07Z")
There are plenty of examples on DL4J here: > **[GitHub - eclipse/deeplearning4j-examples: Deeplearning4j Examples (DL4J, DL4J...](https://github.com/eclipse/deeplearning4j-examples/)** > > Deeplearning4j Examples (DL4J, DL4J Spark, DataVec) - GitHub - eclipse/deeplearning4j-examples: Deeplearning4j Examples (DL4J, DL4J Spark, DataVec) But there is none that is a specific recommendation system. As I’ve said, to build that you’ll need to do some of your own work. In particular you’ll need to learn what it even is in the first place. The way you are phrasing you question suggests that you are in over your head currently. Maybe start with an overview of the topic in general: [What is a Recommendation System? | Data Science | NVIDIA Glossary](https://www.nvidia.com/en-us/glossary/data-science/recommendation-system/)