UNAIR NEWS – Data Science and Technology (TSD) students from the Faculty of Advanced Technology and Multidiscipline (FTMM) Universitas Airlangga (UNAIR) have made another achievement. Tentan Team, consisting of Netri Alia Rahmi, Muhammad Reza Erfit, and Elzandi Irfan Zikra, has secured 2nd place in the Machine Learning Competition at the Data Slayer event held by Telkom Institute of Technology, Friday, Dec 1, 2024, to Sunday, Jan 7, 2024.
The qualifying rounds of the competition were held from Friday, Dec 1, 2023, to Thursday, Dec 14, 2023. During these 14 days, Team Tentan was granted access to 78,500 data specifications of four-wheeled vehicles and their carbon emission levels. The team was challenged to create the best AI model to predict carbon emission levels with various specifications, aiming for the smallest error score possible.
Tentan Team qualified for the semifinals, securing a final model with a Root Mean Square Error (RSME) of approximately 19.05 points. In an interview with the team leader, Netri Alia Rahmi mentioned that if a vehicle has 2,714,100 g/km emissions, Tentan’s model predicted the vehicle’s emissions with outstanding accuracy, differing by only about 19 grams.
Ranked 7 out of 107 teams
On Friday, Dec 15, 2023, the Tentan Team explained their coding results by submitting a notebook in ‘pkl’ format. Tentan Team secured the 7th rank among 107 other teams participating in the competition.
Tentan Team advanced to the final round on Saturday, Dec 23, 2023. In this stage, the team presented their approach, explained insights gained from the data, and provided recommendations for the respective vehicle brands. The final evaluation was based on three criteria: model accuracy, notebook quality, and presentation skills. Tentan Team achieved the 2nd place with a final score of 182.98 points, as announced on Sunday, Jan 7, 2024.
Team challenges
Alia mentioned that during the competition journey, they encountered several challenges. For instance, the received data was highly complex, with the presence of empty data, units in different formats, and diverse data type variations. Data inconsistency and numerous outliers added difficulty to the processing, requiring a considerable amount of time to achieve ‘clean’ data ready for machine learning processing.
The team’s pre–processing addressed empty data, harmonized diverse data units, handled inconsistencies, and identified and addressed outliers. “From this competition, we conducted a lot of research and reviewed journals on carbon dioxide emissions to gain new insights,” she said.
The team then created a new variable called engine efficiency, the ratio of the engine and cylinder variables. Based on various sources, Tentan Team observed that the efficiency of a vehicle’s engine is a crucial factor in CO2 emissions because high engine efficiency can reduce fuel consumption, resulting in lower CO2 emissions.
“And indeed, this variable reduced the error rate of our model,” Alia concluded.
Author: Maryam Fauziah
Editor: Feri Fenoria
