Predictive analysis

[Feras95p]

Dear all,

As you can see in the picture below I have for each number # or row a set of data from Jan to Sep. These values I calculated then by taking the values of the same month in 2020 minus the value of the same month in 2019 divide by the value in 2019 *100%. For example Jan= [Jan2020-Jan2019/Jan2019]*100% . So in this way I can see how much it's increased or decreased as compared to 2019. Now what I'm looking for is to predict  the values of October, November and December based on the existence values. How can I do that?    

[atcodedog05]

Hi @Feras95p ,

Need to use Forecasting model for this.

Being that said its not that straight forward.

[ImadZidan]

Hello @Feras95p ,

My favorite subject.

First of all, you need to layout you data in a correct way. The recommended layout is vertical. so your values will need to be transposed.

Secondly, you will to decide on your null values. you may need to replace with 0. Null values will not work with time series modeling.

Thirdly, your columns need to be date based, as it is monthly, I would say convert to 2020-Month-day. example 2020-01-01.

let me know your thoughts. I will work on a workflow that will give you an idea. But it is worth pointing these things out.

I hope this makes sense.

[atcodedog05]

This is a fun subject.

Really exited to see your solution @ImadZidan : )

[ImadZidan]

hhhhhhhh @atcodedog05 ,

It is indeed exciting topic. However, data isn't exciting at all. Started looking and got a huh.

Lets see how it goes.

[ImadZidan · Accepted solution]

Hello @Feras95p ,

I have looked at the data and I am afraid not much to work with. I wanted to use the TS model factory and TS Forecast Factory to do the bidding but the amount of data is not enough. This would have allowed me to do the grouping of your data.

With your data, I observed that you have an outlier which is the month of jun which was impacting the output so I scaled it down to the maximum of the other records. It is silly but to move forward. 

So, also to move forward hhh, I grouped by month and took the average and built the model on that. This is not to say that you have to do it that way, it was just a way to get the workflow and show you an example.

Well have a look and i hope this will give you an idea.

Disable the Compare model container when you want to run just the final model.