logs

[NetZissou]

2021-05-20

Onboarding Task:

• Completed talent Trac assignments
• Register ADP account

Git/GitHub Configuration:

• Installing Git with the admin help of IT
• Adding GitHub certificate to the WI laptop in order to satisfy the company's firewall protection
• Get the WI laptop configurated to Net's GitHub Account and successfully cloned wicast package repo to the local environment

Research on tidyverts ecosystem: tsibble

• Review the data structure of tsibble, includes
• index, key, interval,
• group_by_key + index_by & summarise combo
• Handling missing values

Research on package development: NAMESPACE

• import function from other packages
• export minimum functions of our packages
• roxygen2 + document() to make tidy changes on NAMESPACE

Some plans in mind for tomorrow:

• Practice research content by adding a new assist function to utils.R, which report the missing information of a tsbl object along with visualization The function will import other functions from ggplot2, pipe symbol from magrittr and etc.

[NetZissou]

2021-05-21

wicast code review:

• After calling check() inside wicast, an error has occurred while generating the examples for functions.
• Located the bug inside of the function add_ci_trace
• Debugging and filed a merge request after the modification Find out an interesting topic while debugging: OO programming in R with S3
• Generic method, the middle man
• Method Dispatch
• Class Inheritance

[NetZissou]

2021-05-25

wicast code review:

• gain some familiarity with the automodel function, but failed to go through the whole S3 class because of lacking directions

Forecasting Analysis Research: prophet

• Saturating Forecasts; caps; minimums
• Changepoints detection and specification
• Customerized seasonality

Some great conversation with Steve about: People end up not using the dashboard products.

• The importance of narrative in data analysis
• People don't want visualizations or dashboards to play around proactively. Instead, they just want to make an action or decision.
• Research on Natural Language Generation. Currently, when talking about NLG, we tend to think of Siri or Alexa. But there is a huge opportunity for NLG to be applied in commercial or corporate settings. From the Google trend data service, we could see that NLP topics are 80 times more active than NLG's. But generally, the hype for NLP is starting to go downwards as well.
• This is an exciting field. Current BI tools are distributing too many fancies functionality for analysts with well-trained data literacies but not for general workers to extract insights. If they go public, consider buying their stocks.

Reviewing the wipeo project:

• Go through detailed documentation and have a clearer image of the business objectives.
• Prompt the suggestions for another way to specify the logistic regression model to interpret the model parameters

Reviewing the CRU forecasting project:

• Research on what is The CRU US Midwest Hot-rolled Coil Index
• CRU being the model response, while commodity price and foreign currencies could serve as the exogenous regressors.
• Follow-up question: could the CRU serve as the market price that we use for the predictions in the wipeo project?

[NetZissou]

2021-05-26

Summarizing work from the past several days. Had a weekly check-in meeting with Wil. More details at here. Some stuff to work on up next:

• How to struct your own S3 objects, or R6, S4
• How to detect seasonalities?
  • Seasonality test could sometimes be inaccurate.
  • Rob Hyndman's post on detecting seasonalities