Eoin: Seasonality cleaned

[K-Ellis]

Take a look at the created_on_year plot for when the cleaned incident data is used in the seasonality study:

https://github.com/K-Ellis/Predicting-Transaction-Times/blob/master/5th%20Iteration/0.%20Results/Kieron/data_understanding/seasonality_study/2017.07.24/Created_On_Year.png

[K-Ellis]

compared to when the uncleaned incident data is used:

https://github.com/K-Ellis/Predicting-Transaction-Times/blob/master/5th%20Iteration/0.%20Results/Kieron/data_understanding/seasonality_study/2017.07.21/Created_On_Year.png

[eoincUCD]

Huh, any idea what's causing it?

While there is a pattern, might not be as bad as it first looks. Can we plot them with the same y axis? Uncleaned data has one outlier messing it up

[K-Ellis]

Not sure.. I'll try deleting the outlier and plotting it again

[K-Ellis]

Here's the uncleaned data:

https://github.com/K-Ellis/Predicting-Transaction-Times/blob/master/5th%20Iteration/0.%20Results/Kieron/data_understanding/seasonality_study/2017.07.25/Created_On_Year.png

And then the cleaned data:

https://github.com/K-Ellis/Predicting-Transaction-Times/blob/master/5th%20Iteration/0.%20Results/Kieron/data_understanding/seasonality_study/2017.07.24/Created_On_Year.png

[K-Ellis]

There's definitely something strange going on.. They look pretty similar up to day ~105, then the variability decreases massively and the last 20 days of data go missing

[eoincUCD]

If we think about the data we have, anything created recently (towards the end of quarter) might not yet be closed out. Is there a filter we are applying that could be causing it?

Maybe park it and test out when we get the next dataset to see if it disappears?