I showed how to login to Celonis EMS using Pycelonis in previous posts. Authentication topic I mentioned there is the most annoying when using API, but after mastering this I can transfer this knowledge to another areas easily.
To master authentication, I would like to show the mechanism of HTTP request in the internet (web) programming. Even I did not know HTTP request well at the beginning of using Celonis, now I got some of basic knowledge and it is enough to use HTTP request.
[Read More]
Limit permissions of API token to minimize risk
At last post Login to Celonis EMS from Jupyter Workbench, I used API key that have same permission as my GUI user. I mentioned that it is too strong and risky against unauthorized access. Imagine your API token accidentally make public, then anyone can operate Celonis instead of you. That is why I segregate API token from Notebook (Notebook may be published to GitHub etc.). Anyway, user API token must be altered to another weaker key especially in production system.
[Read More]
Login to Celonis EMS from Jupyter Workbench
At last post Start Deep Dive to Machine Learning and Action Flow, I introduced overview of these two functions that manipulate Celonis EMS functions without using GUI (browser) for automatation.
From today I would like to share basic functions in Machine Learning (Jupyter Workbench). Before staring I will setup Jupyter Workbench. If you do not have your Workbench, read Share my Analysis by Content-CLI then follow until creating Workbench.
I will go to Launcher in Workbench then select Others > Terminal, then enter below command after $ sign (output is as of 2022-05-14).
[Read More]
Start Deep Dive to Machine Learning and Action Flow
Until last post, I have seen the flow of process discovery through Celonis Process Analytics (or Studio Analysis) and prerequisite ETL (Extraction, Transformation, Load) by Data Integration. I already worked in multiple static process mining projects and found that they are enough functions for static process mining. And other process mining solutions are provided these functions too.
By the way, referring to Process Mining Data Science in Action by Wil van der Aalst, Process Mining can also make it possible for further actions such as monitoring and predictive analysis.
[Read More]
Execute Periodic ETL Automatically
Until last post Validate Data Model by Studio Analysis, I completed creating ETL programs in Data Integration. But this is still test product because this programs are executed manually by operator. In this post, I would like to share the final piece of Data Integration, how to run ETL programs automatically after production release.
There are several ways to achieve automatic ETL, for example in my production system I am using scheduler in Machine Learning Workbench to trigger Jupyter Notebook, then operate Data Integration via Pycelonis (Python API for Celonis EMS).
[Read More]
Validate Data Model by Studio Analysis
At last post Construct My First Data Model, I created Data Model and load data to it. Normally initial load is not perfect, so I should check data in Data Model. Today I would like to share how to validate my Data Model using Analysis. By the way, Celonis EMS main function is now Studio (and App for viewer) and Analysis is also part of Studio, so today I will create Studio instead of Process Analytics to create Analysis.
[Read More]
Construct My First Data Model
Today I would like to create draft version of data model then complete series of ETL (Extraction, Transformation, Load) started from last November.
As I showed in Adjust Time Zone of Event Time in Global Transformation, I created global VIEW for activity table. Before creating Data Model, I would like to create one more global VIEW against case table. As below SQL, it is quite simple to copy all columns of issues in Planio connection, plus one more column _CASE_KEY that is converting from interger id to character.
[Read More]
Inspect Table Data by SELECT statement
Today I would like to share small tips in transformation tasks. Of course main purpose of transforamtion is to create event log table etc. but we are not always sure about source system behavior, so before completing transformation SQL I must investigate data of source system. As I already showed, I can use simple SELECT statement to inspect tables.
Previously when I was familiar with SAP ECC, the only way to easily inspect SAP table is to download table data by SE16 transaction and open Microsoft Excel and simply filter data or use pivot table etc.
[Read More]
Adjust Time Zone of Event Time in Global Transformation
Until now I do not care about value of event time, that is stored at source system like Planio. If I analyse only one system like my current project it is no problem, but if I would like to analyze data from multiple source systems, I should take care time zone of each system.
In my case, I previously analyzed multiple SAP ECC systems, those are located in various area of world (US, EU, JP etc.
[Read More]
Handle Day based Activity as Milestone
I already created four activities until last post Unite SQL statements by CASE Expression, those are fillfilled requirement of event log. Going back to Consider Case ID before Starting Transformation, case ID is the biggest requirement. Also is is not so big as case ID, but event time is important too. In process mining, event time should be year, month, day plus hour, minute, second (YYYY-MM-DD HH:MI:SS in Vertica format). I guess event time in process mining referred to that is recorded automatically by system responding to something action.
[Read More]