timemachines
Capstone data science project
Goals for students
- Adopt a critical, statistical mindset as compare with package-first-learning
- Learn technique for creating incremental, fast, “green” models, as compared with batch-style.
- Acquire skills crucial for meaningful collaboration on open-source projects on GitHub.
Students who may be interested
- Those looking to gain experience deploying, assessing, maintaining and improving models in a real-time setting.
- Those looking to prove to future employers that they are not in Kaggle anymore.
- Those looking to build their open-source reputations.
Suggested steps
- Follow instructions here to join slack and google meets on Fridays.
- Learn GitHub basics by editing the timemachines documentation and making a pull request.
- Gain basic familiarity with using the timemachines package, including installation of constituent packages
- Read the contribution pages.
- Contribute a colab notebook by following the instructions
- Graduate to a good first issue.
- Read up on Python time-series packages list, google around, and locate a promising package or method that is not already included. File an issue by cutting and pasting this issue and making modifications.
Once some familiarity with univariate point forecasting gained, graduate to running distributional algorithms in real-time
- Run the default crawler. See instant participation instructions.
- Get familiar with predicting using python.
- Use StreamSkater instructions to launch a crawler that uses a model in the timemachines package.
In-person help
See step one.
-+-
Documentation map
