Nashua High School Senior Selected In National Science Competition

Jan 18, 2019

Credit courtesy of Tejas Sathyamurthi

A high school student from Nashua has been selected as a scholar in a national academic competition. 

The Regeneron Science Talent Search is one of the nation's oldest and most prestitgious math and science competitions for high school students.

Tejas Sathyamurthi, a senior at Nashua High School South, was chosen for his project on using machine learning to predict forest fires. And Morning Edition Host Rick Ganley spoke with him about his win.

(Editor's note: this transcript has been edited lightly for clarity.)

Can you tell us about your winning project?

Yeah, so essentially what I had worked on was using machine learning to analyze data from preexisting fires available on the Internet, and then use that to analyze and be able to predict with a certain accuracy fires that occur in the future. Because of my past inspiration, and the growth of technology has been so rapid over the past couple of years, and I think the evolution of computers -- it was a reason why I decided to choose machine learning and computers.

This is a fairly new science, isn't it? It's really made possible by big data and fast computing.

Definitely, yeah.

So how did you get interested in that?

So I'd say I was inspired because on a visit to California in 2015 I witnessed a small fire on the side of the road. And coming from New England that's something I'd never seen before. You know I'm so used to snow [and] ice, those kind of storms. But that was something that really hit me, because even though it was something small, I realized because of how vast the land was. If it was on a larger scale, it probably would have eradicated the whole land and possibly killed more people, as we've witnessed with the Woolsey Fire unfortunately in the fall of 2018. And then I think that situation is probably the spark that led me to work on this project.

So how did you put this all together? You went about collecting this data. You said that this data was publicly available?

So yeah, machine learning essentially works like you teach a computer to do something. And basically I used this data set to teach the computer the different instances and the combination of factors that lead to a fire. And in order to test it, I created my own data set compiling data from Weather Underground on relative humidity, and then using online historical data on the major fires that have occurred over the past 50 to 100 years in the United States. So once I compiled all this, I tested my model and the data for the United States and the accuracy came out to be approximately 76 percent.

This sounds a lot to me like weather modeling. But it sounds like your accuracy rate is right up there with some of the best weather modeling that's available.


I can't imagine the hours that went into this project. How did you do that and balance your schoolwork and other activities in life?

Throughout high school one thing that I've definitely learned more of is time management with all the courses and advanced classes that I've been taking. I began this in 2016 after the summer when I traveled to California. And I think it was just something that because I was like passion about, I was eager to work on it, go through those background studies and learn more about it.

How did you find out about the win? I'm wondering what that felt like.

Yeah, so I didn't know. Like my mom was the one who gave me the call and said that I'd won. I mean it was really exciting, and like I was just so grateful for this.

I know part of your reward for the competition is $2,000 that will go toward future education. What are your plans for after graduation?

Yeah, this fall I'll be going to college. And I think based on my past research and everything I'm planning on majoring either in statistical data analysis/data mining because that's an area that's definitely had a major impact on me. And I've had a passion for math and numbers, and hopefully in the future I'll be able to join the workforce as an analyst in like statistics or a risk analyst.