After Week Nine in Seattle: Programming Partner Problems
Aaron W Chen here again with a #vlog from #Seattle! If you've seen my videos before, welcome back! If not, I'm a photographer and chemical engineer turned data scientist and I moved to Seattle to attend Flatiron School's Data Science Bootcamp!
We start with a review of my favorite photos from the previous week. For photo 1, here's an abstract photo of stacked and torn posters. For photo 2, here's a leaf imprint in concrete! For photo 3, here's a composite of Monorail Espresso. Photo 4 is of a leaf from Teneriffe. Photo 5 is of leaves of different colors on rocks. Photo 6 (my favorite) is of leaves above.
Let me know what you think! Would you pick different photos?
What happened this week at coding bootcamp! This week was a project week for Module 4? We worked in pairs to analyze Zillow data ending in 2018 to predict which were the five best zipcodes to invest in for 2019. The project itself was cool: we had to apply time series forecasting and critical thinking about the problem before we doing anything. Unfortunately, I did not work very well with my partner.
The most common ways of handling time series forecasting are Autoregressive Integrated Moving Average Model (or ARIMA) https://machinelearningmastery.com/ar... and now Facebook's Prophet https://facebook.github.io/prophet/. Each has it's pluses and minuses, and I encourage you to read about both, but one thing I picked up pretty quickly was that ARIMA requires custom tuning for each data set that you have. Prophet is highly automated and gives good to excellent results, but tweaking ARIMA can provide better results than Prophet https://blog.exploratory.io/is-prophe...
However, our data set had something like 15,000 zipcodes and thus 15,000 custom models if we didn't filter down. I did some rudimentary filtering based on traditionally famous real estate markets and made it clear that ARIMA was not going to work. The Dallas/Ft Worth area alone had something like 200 zipcodes in our dataset.
I asked my partner to switch over to Prophet so we could actually start really thinking about zipcodes and doing analysis. He refused and insisted on using ARIMA. At that point, I realized that we'd be working independently and mash our work together for slides.
Our data ended in 2018, which meant that Zillow would have 2019 data for my models to compare to! I figured that companies specializing in analyzing this data would have interesting articles discussing the market, so I started reading articles about hot markets in 2018 and 2019. This reduced the number of zipcodes I was going to look at. The ones I decided to focus on were referenced in Zillow's article talking about the areas with the highest search traffic from both inside and outside the metropolitan area combined with 2019 market info.
However, as someone who lives in San Jose, I figured something was not entirely correct. In 2019, Bay Area house prices FELL 10% on average but the Zillow projections estimated an increase.
I eventually ended up with a few metropolitan areas ("Currently Trendy Zipcodes of Interest") for Prophet! I compiled pricing data for those areas so I could see how the model did. In addition, I was aware that the yield curve had inverted and wanted to use that information. My recommendation was NOT to invest in real estate as these areas appear overvalued and will suffer should another recession come, but I continued doing the project as if our client insisted on real estate.
I got a new camera and lens! The GX9 with 20mm 1.7. I'm liking it so far! It isn't as fast or good as my G9, but it does fit in my pocket!
It was a Strength week in my workout program, so my routine had me do squats, bench and military presses, and deadlifts.
Please thumbs up, thumbs down, comment, and share!. Also reach out to me on Facebook, Twitter, Instagram. Lastly, do consider subscribing!
Facebook: / composedandfocused
Instagram: / aaronwchen
Twitter: / cofostudios
Below are some Adorama affiliate links!
Panasonic G9: https://adorama.rfvk.net/PgqbM
Panasonic Lumix DG Leica Vario-Elmarit 8-18mm f/2.8-4 Lens: https://adorama.rfvk.net/qvGyb
Panasonic Lumix G Leica DG Vario-Elmarit 12-60mm F/2.8-4 Power O.I.S. Lens: https://adorama.rfvk.net/kvPWx
Panasonic Lumix G Leica DG Vario-Elmarit 50-200mm F/2.8-4 Power O.I.S. Lens: https://adorama.rfvk.net/L1Y2V
Panasonic DMW-TC20 2X Lumix Teleconverter: https://adorama.rfvk.net/DmM35
Rode Microphones VideoMic Pro+: https://adorama.rfvk.net/EgBAe
Tascam DR-10L Digital Recorder: https://adorama.rfvk.net/KGV2N
Sony Cyber-shot RX0 II Digital Camera: https://adorama.rfvk.net/JvmEE
SmallRig Cage for Sony RX0 II Camera: https://adorama.rfvk.net/ry1Xj
Zhiyun Crane M2: https://adorama.rfvk.net/OZXLG
DJI Mavic Pro Fly More Combo: https://adorama.rfvk.net/kvPM0