Vietnam is the world’s second-largest coffee exporter and accounts for more than half of the global Robusta supply. Coffee production in the 2022/23 crop year is expected to reach 29.75 million bags, of which Robusta accounts for more than 95%.
In the International Coffee Organization's 2021/2022 Annual Review, Vietnam ranked first in coffee cultivation productivity with 2.4 tons/ha. Coffee production in Vietnam is made up of Robusta, Arabica, Cherri, Moka and Culi beans, which are the most popular coffee beans grown in Vietnam.
However, prices of agricultural products in general and coffee bean prices in particular are often unstable and can fluctuate sharply during bumper harvests, significantly affecting farmers' incomes and causing damage to the economy.
From left to right: Students of RMIT Faculty of Science, Engineering and Technology: Lam Tin Dieu, Nguyen Hai Minh Trang, Nguyen Phuong Nam (top row), Le Ngoc Nguyen Thuan, Doan Chanh Thong (bottom row)
From left to right: Students of RMIT Faculty of Science, Engineering and Technology: Lam Tin Dieu, Nguyen Hai Minh Trang, Nguyen Phuong Nam (top row), Le Ngoc Nguyen Thuan, Doan Chanh Thong (bottom row)
To research a solution to this problem, over a period of four months, a group of final-year students majoring in Information Technology, Faculty of Science, Engineering and Technology, including Nguyen Hai Minh Trang, Doan Chanh Thong, Le Ngoc Nguyen Thuan, Nguyen Phuong Nam and Lam Tin Dieu, trained and evaluated six machine learning (ML) models to predict coffee prices, which can help Vietnamese farmers make informed decisions about their crops and plan accordingly, optimizing profits and minimizing losses.
“We developed six ML models, namely LSTM, GRU, ARIMA, SARIMA, SVM and RF, based on the history of coffee prices, gasoline prices, temperature and rainfall, to predict Robusta coffee prices in Lam Dong province and found that the RF model, using the entire dataset, was the most effective,” Trang said.
Among the 6 machine learning models, the RF model, using the entire dataset, produced the best results.
“RF can incorporate richer datasets and handle nonlinear relationships. Additionally, fuel price proved to be a significant predictor and outperformed all other tested features combined.”
The team stressed that the model has potential for further improvement by studying and incorporating the impact of crop yields, market trends and geopolitical events on agricultural prices.
Each team member faced different challenges during the project, such as lacking in-depth understanding of different ML models, effectively communicating the complexity of what they were doing to the AI domain, or managing time and communication when working remotely. However, by investing significant time in research, digging into AI and ML-related research papers, and improving their technical and collaboration skills, they improved their AI research skills for real-world problems and were able to develop their team’s research into real-world products.
“The main challenge for us revolved around data collection and integration,” Thuan shared.
“While developing the model was fairly straightforward, the significant time required to collect and combine the data posed a huge challenge for us. Each team member went through a learning curve and advanced their skills in both technical and project coordination, from in-depth research, to pushing innovation and coming up with new solutions.”
At the time of the study, Nam was working from Hanoi and had a full-time job. To prevent delays and potential disruptions, Nam said the team set up weekly meetings and maintained regular communication, both to motivate each other to stay on track and to complete the assigned workload.
The team’s capstone project was closely supervised by faculty from the School of Science, Engineering and Technology, RMIT Vietnam. The project results were recently presented at a prestigious international event – the 8th IEEE/ACIS International Conference on Big Data, Cloud Computing and Data Science Engineering (BCD 2023) – with researchers, scientists, engineers, and experts in the fields of Big Data, Cloud Computing and Data Science.
Student Nguyen Phuong Nam demonstrates how the coffee price simulation website works
The team plans to refine the models based on feedback from the conference presentations, and also explore other approaches to improve the accuracy and applicability of their predictions.
“We plan to delve deeper into cutting-edge techniques and emerging methods in the field to further strengthen the research results the team has achieved,” Thong said.
“In addition, we plan to collaborate with other experts in the field and explore potential partnerships to expand the scope and impact of the group's research findings.”
The team plans to continue to iterate and upgrade the research so that it can make practical contributions to the ever-evolving field of Big Data and AI from your specific research.
Source link
Comment (0)