Making contributions to the broader analytics community is fun benefit of Big Lake Data’s on-going investment in professional development. As such, we’re honored to have worked with J.A. Marin and Stephan Zohren in the development of an R software package for micro-climate analysis.
If you are an R-user, we invite you to view the basic README below and explore the development version of this software yourself. Your feedback is welcomed!
wundr provides API interfaces to Personal Weather Station (PWS) data maintained by Weather Underground. Tables of PWS locations and metadata for user specified geographic areas are constructed. Robust retrieval of current and historical weather condition data for selected PWS. Computational tools for spatial analysis, including kriging (Gaussian processes) for interpolation, as well as prediction of missing data and forecasts. Visualization of micro-climate data methods including contextual static maps and interactive web maps.
This package was developed as a final project for Prof. Balasubramanian Narasimhan’s Paradigms for Computing with Data class at Stanford University.
Install the development version from github
# install.packages("devtools") install_github("MatthewSchumwinger/wundr", build_vignettes = TRUE)
Key features of wundr include:
- Creation of S4-Class Table describing a given region’s PWS.
- Subsettable tables based on #1 above.
- Robust web retrieval of data and storage in memory.
- Visualization of micro-climate data through density maps and time-series.
- Computational tools for spatial analysis, including kriging (Gaussian processes) for interpolation, as well as prediction of missing data and forecasts.
- Web interface for interactive visuals using leaflet.js.
To see examples of these features in use, please view the ‘overview’ vignette.
vignette(package = "wundr")
We have a professional development budget here at Big Lake Data. As such, I recently completed a graduate course in data mining at Stanford University. The best part of the course was competing as a team with two other working-professionals on a predictive modeling contest sponsored by Kaggle. We programmed everything in R. And for a guy like me, who hacked my way into R, collaborating with a couple of wicked-smart computer scientists was a revelation.
We won. You can read the hoary details of how we did it from the Kaggle winners’ blog. No cash prize, but substantial bragging rights awarded . . . and claimed.
Predictive modeling? Other types of analytics? Yes, we can help you with that. Get in touch.
Sometimes beauty is inherent. Sometimes it’s familiar. And sometimes it’s both.
For me, John Cheshire’s visualization of the world’s population is both. First, there’s an elegance and economy by which the data is encoded. Then, there’s the striking similarity between his map and Joy Division’s Unknown Pleasures iconic album cover, which featured a plot of pulsar radio signals.
If you’re a fan of British post-punk rock and Edward Tufte – you’ll probably find Cheshire’s work beautiful in both ways. Heck, you can even order a print of his world map and hang it on the wall right next to your stacks of vinyl.
(As a bonus for R users, Ryan Brideau has a neat post on how he reverse-engineered the visualization and applied it to New Brunswick.)