Leading Lights: How teamwork and the will to learn helped two HK students win the AI Challenge: Weather Forecasting Competition

Leading Lights: How teamwork and the will to learn helped two HK students win the AI Challenge: Weather Forecasting Competition

We talk to the winners of a weather forecasting competition, who outsmarted 30 other teams with their AI-driven temperature predictions

24bb6206-18b4-11e9-8ff8-c80f5203e5c9imagehires132304.jpg

TWGHs Mrs Wu York Yu Memorial College students Ken Kong (left) and Jacky Ye won the Artificial Intelligence Weather Forecasting Competition.
Photo: Joanne Ma/SCMP

The idea of artificial intelligence is no longer wild and futuristic – it’s an increasingly integral part of our lives.

“I think AI technology will bring a lot of convenience. From self-driving cars to the spam-filtering function in your email inbox, AI is involved,” said Jacky Ye Rongjie, a student at Tung Wah Group of Hospitals Mrs Wu York Yu Memorial College.

Last November, Jacky, 17, and his classmate, Ken Kong Hok-ming, 16, won the AI Challenge – Weather Forecasting Competition, which was co-organised by Hong Kong Observatory (HKO) and the Hong Kong Meteorological Society.

Chang'e 4 landing on dark side of the moon made possible thanks to tech and decisions made by HK PolyU researchers; professor involved tells students to ‘dare to explore’

Thirty-one teams from 24 secondary schools took part. Participants had to use machine learning (an algorithm that allows machines to become more accurate in predicting outcomes on their own) in AI, and meteorological data from the past 10 years provided by HKO, to predict the temperature at the Zero Carbon Building in Kowloon Bay from November 26 to 30.

Over the five days, they submitted the hourly temperature forecast from noon to 11pm at the Zero Carbon Building for each day, 24 hours in advance. In other words, they had to generate 60 predictions in total. Whichever team was able to predict the most accurate temperatures, best demonstrate their application of machine learning, and show the best teamwork, would win.

Jacky and Ken beat La Salle College, who came second, and PLK Tang Yuk Tien College who came in third. Seven other teams received merit awards.

How alternative environmental solutions like desalination plants and wind turbines aren't really fixing our planet's problems

“I had very little knowledge about AI before, and now I’ve learned so much about AI techniques, as well as the importance of teamwork,” Ken said.

“Jacky helped me out a lot, though, and I think we complement each other very well.” The duo divided the work between them based on their strengths. Jacky, who’s more familiar with AI technology, was responsible for optimising the algorithm and getting accurate data, while Ken was in charge of research, gathering the data, and finding patterns in the data.

“We used a programming system called Random Forest, and what we needed to do was train the algorithm based on the set of data we’d been provided with.”

La Salle College student crowned winner of Hong Kong maths competition

Each team’s algorithm was slightly different so they each produced different data. The method used to gather, analyse and present the data varied between the participants, too.

The algorithm had to be able to process a big set of data, and work out the relationship between different pieces of information. Jacky and Ken trained theirs to figure out how data on wind speed, humidity and time of day would affect the temperature.

Jacky and Ken used Python to constructed the algorithm for this competition.
Photo: Joanne Ma/SCMP

The pair divided the data provided into two sets. One was used to train the AI, while the other was used to check their own work in the trials. They used their AI to predict the weather in the past, and compared the results with the real-life data, to check if their algorithm was producing accurate results.

“The results weren’t accurate enough in the beginning, because the weather could be affected by so many factors, and our algorithm wasn’t complete yet,” said Jacky. But they were able to get more accurate results once they factored in all the data they had that would affect the temperature.

HK students who built a Rube Goldberg Machine on how it taught them to be better listeners, teammates, and inventors

The two Information and Communication Technology students will be sitting the HKDSE next year. After that, they’re both planning to apply for programmes related to AI or computer engineering at Hong Kong University of Science and Technology.

“I’ve always been interested in robots since I was a kid. Ever since I realised robots have their own intelligence … I was intrigued, and really wanted to do something with it,” said Ken.

Edited by Nicole Moraleda

This article appeared in the Young Post print edition as
Predicting the weather with AI

Comments

To post comments please
register or