What spiega is about
- Knowledge sharing
- Portfolio
Time series forecast
delivery pipeline
This readme explains the content of the directory and the execution steps to complete the delivery forcasts.
delivery curves
This project folder does not contain any script to retune or re perform a mapping or a prediction or run sanity checks, it just applies already pretrained models. In case of unconsistency please refer to the author.
setup
In the project directory there is an example docker file for creating an ad hoc container.
project directory
To install the basic libraries please run this script with python 3.6+. Depending on the required KPIs other graphic libraries might be installed.
All the scripts consider the variable $LAV_DIR which can be set as the current directory. Paths can be otherwise hard coded adjusted.
tdg/cronon production
jupyter notebook on Hungarian cluster
- check data availability for tripEx-act and adjust dates in the qsm job for activities
- check data availability for analyzer and adjust dates in the qsm job for odm_via
- get permission from QA about submitting the jobs and set priority
- schedule the jobs
- check if the job run without any error on the server
- use the following notebook to postprocess the data - activity
- use the following notebook to postprocess the data - odm_via
- download the two files from the jupyter cluster
local production
example of mapping result
- Complete the date list in this file
-
apply this script to dowload the weather information for training
-
apply mapping with this script
- copy the cilac actRep in the directory
-
the output will be written write to act_weighted
-
apply prediction with this script on
- the prediction will be written in act_predict
results after regression
KPI
differences wrt previous delivery
Apply this script to sum up the results for the customer delivery and check the following KPIs:
- maunal visualization of activities and via counts for all locations
- consistency with previous month
- consistency with seasonal trends
- spikes in data
- expected range in capture rate
- consistency in ranking
- comment events that can influence the specific date/location
t-test feb-mar vs t-test mar-apr, change in ranking
internal meeting
differences in capture rate between deliveries
- organize an internal meeting at least two days before the customer delivery
- present the results against the agreed KPIs
- show a presentation similar to this former delivery
- in case of quality issues:
- assign tickets to the responsibles
- reschedule the analysis (job, processing...
- reschedule the delivery
check against isocalender
finalize the delivery
Run the script tank_delivery adjusting the output file name.
finalize the delivery
- manual check the numbers and prepare the excel file like in the former delivery (pivot...)
- update BI dashboard (spotfire - tableau)
customer support
- be available on customer feedbacks and invite the concerned responsible to a meeting
- assign tickets/solve issues
- reschedule the delivery
- update the delivery