D607_P03_presentation_DRAFT

What are the most valued data science skills?

Bar Raisers - Large Group Justified

  • Bharani Nittala
  • Richard Sughrue
  • LeTicia Cancel
  • George Cruz Deschamps
  • Jack Wright

For more details on authoring R presentations please visit https://support.rstudio.com/hc/en-us/articles/200486468.

Misc Notes:

Cite Slide (Letter) for:

  • text inclusion/revision
  • code snippet inclusion
  • plot image inclusion - do 2 things give me exact name of image

For the plot image inclusion, do the following:

  • NameTheReference of your plot image within the brackets
  • plot image file name within the parenthesis

NameTheReference image

Example: quanteda image

SAMPLE Visualization(s) - using exported plots

Left Column Header

  • text above image trythis image
  • text below image

Right Column Header

  • text above image trythis image
  • text below image

(Slide A) Overview

  • What are the most valued data science skills?
  • CIO Perspective
  • Webscraping from Social Media Sources
  • Tidying, Transforming, and Storage
  • Exploratory Data Analysis
  • Key Themes from Social Media Sources

  • Summary / Conclusions

  • Future Exploratory Data Analysis Opportunities

  • Questions

(Slide B) What are the most valued data science skills?

This project used an exploratory data analysis approach to obtain information available in the public domain from leading social media websites to answer this question.

Websites Scraped:

CRISPER CRoss-Industry Standard Process for Data Mining CRISPER image

(Slide C) CIO Magazine - Bob Violino, Contributing Writer & CIO

Essential skills and traits of elite data scientists are -

  • Critical thinking
  • Coding
  • Math
  • Machine learning, deep learning, AI
  • Communication
  • Data architecture
  • Risk analysis, process improvement, systems engineering
  • Problem solving and good business intuition

Reference: Violino, B. (2018, Mar 27). Essential skills and traits of elite data scientists. CIO Magazine

(Slide D) Data Extraction through Web Scraping Methods

  • Sources: LinkedIn, Twitter, Amazon Jobs, Glassdoor
  • Searched and scraped datasets for “Data Science”
  • R Packages used: tidyvest, rvest, httr, jsonlite [name others]
  • GadgetSelector (not sure - maybe LeTicia used??)
  • “Inspect Element” feature within browser

Code snippet:

  • Initially scraped all jobs listed within NYC metropolitan area
# base url for all jobs search
base_url<-"https://www.linkedin.com/jobs/search/?geoId=90000070&location=New%20York%20City%20Metropolitan%20Area&start="
max<-30
#scrape linkedin
df_null_linkedIn<-linkedIn_scrape_unique(base_url,max)

df_null_linkedIn%>%
  filter(unique())

(Slide E) Tidying, Transforming, and Storage

Methods of Tidying & Transforming -

(Slide F) Tidying, Transforming, and Storage - Quanteda

Quanteda

  • Quantitative Analysis of Textual Data
  • Takes “text mining” to the next level
  • Key Concepts - corpus structure, tokens, stopwords

quanteda image

jobs_vector<-as.vector(df_data$description)

job_corpus<-corpus(jobs_vector)

dfmat_data<-dfm(job_corpus,
                remove = stopwords("english"), 
                stem = FALSE, remove_punct = TRUE
)

(Slide G) Exploratory Data Analysis Methods

  • Explain high level tasks performed here - initial observations/impressions

Preview of visualizations: LinkedIn-A image trythis image

Twitter-B image trythis image

(Slide H) Visualization(s) - Twitter with LinkedIn scrapes

LinkedIn

  • some text above image LinkedIn-A image
  • some text below image
  • specific characteristics expressed more often for DS jobs than non-DS
  • some more text

Twitter

  • based on twitter #hashtags Twitter-A image

  • some text below image

  • other points

(Slide I) Visualization(s) - Twitter

  • Extraction snap shot of most recent 1,000 “DataScience” tweets at time of execution

  • Ranked bar graphs of #hashtags embedded within extracted tweets Twitter-B image

  • Elaborate on key info features of EDA from Twitter

# put excerpts of code sets here -

(Slide J) Visualization(s) - LinkedIn

Wordcloud LinkedIn-Wordclound image

  • other points

  • Elaborate on key info features of EDA from LinkedIn

# put excerpts of code sets here

(Slide K) Visualization(s) - Salary ranges GlassDoor

Glassdoor image

  • time frame represented in extraction:

  • other points

  • Elaborate on key info features of EDA from GlassDoor

# put excerpts of code sets here

(Slide L) Visualization(s) - Amazon Jobs in Data Science

AmazonJobs image

  • time frame represented in extraction:

  • other points

  • Elaborate on key info features of EDA from Amazon Jobs

# put excerpts of code sets here

(Slide M) Visualization(s) - R versus Python

According to Anand Rao, global artificial intelligence and innovation lead for data and analytics at consulting firm PwC (Violino, 2018):

“language of choice in data science is moving towards Python, with a substantial following for R as well”

The evidence from the following graph affirms this point.

RvPython image

# put excerpts of code sets here

(Slide N) Visualization(s) - Data Scientist Internships

One observation from data extraction is the number of Data Scientist Internship positions with:

  • Spotify
  • CVS Health
  • others identified in LinkedIn extraction
  • Amazon Jobs list ??

Internships image

# put excerpts of code sets here

(Slide O) Summary/Conclusions

So What are the most valued data science skills?

[Pick following from Violino's list affirming points in his 2018 article based on our EDA]

  • Critical thinking
  • Coding
  • Math
  • Machine learning, deep learning, AI
  • Communication
  • Data architecture
  • Risk analysis, process improvement, systems engineering
  • Problem solving and good business intuition

(Slide P) Future Exploratory Data Analysis Opportunities

Enhancing this framework to consider questions such as:

  • Increase in working remotely in DS as a result on COVID-19
  • Association between level of compensation and years of experience / higher learning
  • Influence of MOOC (Massive Open Online Courses) with increasing learning opportunities
  • YouTube as learning channels (> 100K DS videos loaded since beginning 2020!)
  • Certain industries investing more heavily in DS?

Questions -