The wildfire whisperer: Albert Um
A new entrant to the data science field wins hackathon with unorthodox approach
Albert Um, Analyst at Horizon Media, is a data science practitioner that began his data career with the IBM Data Science Professional Certificate through Coursera. He was previously in the manufacturing industry, assembling women’s outerwear for contemporary brands in New York City. Shortly after completing the Coursera course, he enrolled in a data science bootcamp, Flatiron School, to further develop his newly found passion.
What was your motivation for getting involved in the Call for Code Wildfire Prediction hackathon and how did you approach coming up with a model?
After graduating from the bootcamp, I was actively looking for employment. At the same time, I didn’t want to stagnate my learning and searched for a community to aid my development. I stumbled across the Call for Code hackathon while scrolling through my LinkedIn Newsfeed. I was immediately interested because I wanted to apply/learn newer sequence models and the wildfires Spot Challenge seemed like a good exercise. I was interested in utilizing a Convoluted Neural Network(CNN) into a time-series problem. Traditionally, CNN models are used for image classifications however Convoluted Layers can be applied to sequence problems and are generally faster than Long Short-Term Memory (LSTM) or Gated Recurrent Unit (GRU) layers.
After taking part in the wildfire hackathon, what role do you think data science can play in dealing with wildfires and other issues we face?
In short, I believe data science can solve wildfire disasters. I think wildfires only become disasters once the expansions of fire start to outpace the capacity of the firefighters containing the spread. And, I suspect decision-makers can prioritize certain areas over others with some intuition and experience. For example, let’s say there are two areas (of 1km area pixel) on fire. One is surrounded by a body of water, while the other is surrounded by dry grasslands.
If both areas are of equal importance (same landscape details), then I might intuitively prioritize the fire surrounded by dry grassland in fear of spread.
Unfortunately, analyzing fires at a 1km granularity will result in hundreds of petabytes of data.
The Australian Wildfires Spot Challenge prepared the data by aggregating the estimated fire areas by region on a daily basis and was more manageable for this sprint. In further studies, I believe data science will play a major part in dealing with wildfires.
What skills do you think are most necessary to to make a difference with tech for good?
I think ‘Documentation’ is the most necessary skill to make a difference to use tech for good. Making your work replicable for others will play a more significant part in ‘tech for good’ than individualistic powers.
The programming and statistical skills required are mere necessities that everyone will need to overcome. I believe meaningful impacts are born due to the people who have laid the foundation for the next wave of pioneers.
Feeling inspired to drive positive outcomes for the world through technology? Check out the projects and get involved today.
If you liked the story, be sure to give it a clap and follow Call for Code Digest for more tech-for-good stories! Also, receive monthly updates on the Call for Code challenges, coding resources, meetups, and more, straight to your inbox!
