Help with Forecasting

[dlopez]

Hi Community, 

I'm trying to produce some aggregated forecast and was in need of some help, mainly around how to summarize the data before passing it through the TS models. The data is at the employee level but we'd like to have it summarized to the dept ID level. Below is my problem statement, and then a desired outcome. The dataset is also attached. 

Problem Statement: We'd like to forecast out attrition in our org for the next three months by dept ID. 

Desired Outcome: two columns, first column would have dept ID and the second column would have the forecasted attrition.

I know I need to summarize, but I imagine I need to summarize by date, and then add by Dept ID? Any help would be great. 

Thank you. 

[RobertOdera]

Here you go, @dlopez 

David, I thought it best to comment on the workflow (in hoes that, if you find value, you could have it all in one place).

Please like + mark as an acceptable solution if you find the approach + comments meaningful!

Cheers.

[dlopez]

Hey Robert, 

Thanks for all the work there, great dialogue going on here and I greatly appreciated. 

The work you've done is a general forecast which is great, but doesn't really meet the need I'm working towards, which is to forecast by dept which means forecasting by group. I know you mentioned that you're not the biggest fan, but it's what I'm trying to achieve and is why I used the TS Model/Forecast Factory. I'm not quite sure what you meant by "Forecast without grouping, and then ALLOCATE the forecasted number to Dept ID based on EDA mix", are you talking about doing some EDA to better understand population segmentation (focus on the populations where you see turnover) and then run those groups through a TS model? 

Thanks again for the dialogue, it's been great. 

Cheers, 

david

[RobertOdera]

Sure thing, @dlopez 

The focus should always be on Outcome versus Process, so go for it - forecast by Dept ID.

However, kindly also consider two things:

• Segmentation + Clustering on the cleaned data set and then forecast by Cluster
• As I previously described, which is a generalized first-blush approach (give Simple a chance)

-What I meant by forecasting without Grouping = forecast by overall attrition versus by each Dept ID, so:

1. Let's say you have two departments, 12 and 13, and your "EDA-ish" process shows that Normalized attrition mixes out to a ratio of 65:25 for the departments.
2. Let's also say that your absolute point number from the overal forecast is 100 for the forecasted period.
3. You will allocate the 100 across the departments as 65 for Dept 12 and 25 for Dept 13

*I'm curious as to whether the Dept ID are meaningfully different (for example, one department = Janitorial and another = HR) because that would suggest a meaningfully sufficient differentiation as to warrant seperate forecast streams (perhaps industry turn-over factors and magnitudes are hugely diffrent across those two human asset sectors)

**Still, first = let's help you as we can according to your current approach

***Next or second = stay open to exploring what else is possible

Either way, happy to help as I can + get on a video call to chase that Rabbit!

Cheers.