leaflettuce / teradata_challenge Goto Github PK

• remove null cols (done)
• set data types
• save new datasets into interim dir

Get rough idea of the difference in hire likelihood and time to hire based upon demographic data.

And Clean as needed.

Unemp rates by state?
Jobs epr industry in states?

• charity nav ranking?
• description of demo
• state distr
• edu / disab (inequalities)
• states deep dive
• branch and rank trends
• likelihood and time by volunteer
• volunteer (who)
• visualize log reg

From rough slides Chris maps together lets review and finalize then gtfo 😁

Split out from #3, we need to figure out how to track time to hire to answer business problem revolving around demo differences and hiring time #2.

As @mitchb63 mentioned, there are time_to_(color) variables in df_contact we can use for this.

Should time to blue be the dependent for this question? Or should we create one based on date user account was created to date turned blue? OR some other method?

Determine focus for the project and analysis.

RESOURCE: business_questions.txt in main dir.

For updated issues and tasks for this project, see Kanban taskboard.
Will not be update everything here. (Only code related tasks).

Once we get a grasp of the var's we need from #7, we could drop any unneeded columns from the tables to make working with them easier (and quicker) in Tableau/PBI/Py?etc...

I think we have a good start to the client demo and time in system business questions in #2 .

Does anyone see a way to identify which clients had volunteer assistance? Or any other way we can go about figuring volunteer 'effectiveness'? I assumed it would be a var in contact but I can't seem to find it. Any thoughts?

Census.gov and VA.gov are solid sources for this.

Use summary skills 'qualifications' variable to map out a network diagram.

steps

• filter out to only skills and hired columns
• split by hired/not hired (maybe?)
• create co-occurance matrix
• use networkx to map
• aesthetics
• EZ, right andy 😬

Once data types are set from #1, does any text within the data need cleaned?
This should probably only be done on data/variables that chosen business questions will require to save time.

For example:

• input errors such as 'lieutenant' v 'leutenent' (misspellings)
• remove unneeded sections, '01 campaign advisor' (remove 01)
• splitting values, 'LA - March' into 'LA' and 'March'

Create word cloud of summary skills - hired vs. not hired.

It could be beneficial to get a list of the variables that will be used to solve the business question listed at #2

Anyone have thoughts on best route for this? We could keep a simple txt doc in the main dir listing them out, or add them to the EDA summary excel sheet maybe?

Once we have this it might be worth creating a custom dataset or two that drops all columns aside from these of interest to keep analysis less complicated.

set each client in contacts to an 'service era' based on standards provided by HH.

Once questions are decided #2 and cleaning done #3 we should join the datasets as needed for analysis.

GG everybody!

Narrow focus of how to solve business questions. Though sharing EDA and discussion best route for presentation. To occur 3/30.

create var based on time from date to green to date turned blue.

RESOURCE
https://www.teradatauniversitynetwork.com/Teradata.University/files/8f/8f94d19d-c099-4fc3-9ea7-1ee85112655c.pdf