Prediction and Identification of Key Factors Associated with High Soft Drink Consumption Among University Students Using Machine Learning
This study utilized machine learning, specifically the Random Forest algorithm, to analyze data from 450 university students and identify smoking status and academic year as key predictors of high soft drink consumption, achieving 77.04% accuracy to inform targeted health interventions.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the crowded aisles of modern life, sugary beverages have become a constant companion, particularly for young adults navigating the transition to independence. These drinks, often carbonated and packed with refined sugar or artificial sweeteners, offer a quick burst of energy and flavor, but health experts have long warned that their liquid calories do not satisfy hunger the way solid food does. This disconnect can lead to a gradual increase in overall energy intake, contributing to weight gain and a host of other health issues like dental decay and metabolic disorders. While the risks are well-known, understanding exactly who drinks the most and why remains a complex puzzle. People make these choices based on a tangled web of personal habits, social pressures, and what they believe about the drinks themselves. To untangle this web, researchers are increasingly turning to a powerful new tool: machine learning. Unlike traditional methods that look for simple, straight-line connections between two things, these computer algorithms can sift through vast amounts of information to find hidden, non-linear patterns, revealing which factors truly drive behavior in ways that human intuition might miss.
A team of researchers at Mawlana Bhashani Science and Technology University in Bangladesh decided to apply this advanced approach to a specific group: university students. They wanted to move beyond simple surveys to build a predictive model that could identify which students were likely to consume high amounts of soft drinks and, more importantly, understand the specific traits that made them so. The team gathered data from 450 undergraduate students across the campus, interviewing them face-to-face to gather details about their lives. They asked about basic demographics like age and year of study, but also delved into personal habits such as smoking and alcohol use, and even their attitudes toward the drinks themselves, such as whether they believed soft drinks were tasty or if they thought they contained harmful preservatives. The researchers defined a "high consumer" as anyone drinking at least one liter of soft drinks per week, a threshold that allowed them to clearly separate heavy drinkers from the rest of the student body.
The initial look at the data revealed a striking picture. Roughly one-third of the students, or 33.6 percent, fell into the high-consumption category. When the researchers began to compare the habits of these heavy drinkers against those who drank less, clear patterns emerged. The students who drank the most were significantly more likely to be smokers and to consume alcohol. Their attitudes also differed; heavy drinkers were more likely to agree that soft drinks were very tasty and less likely to believe that the drinks contained harmful ingredients or that they caused health problems. Interestingly, the year a student was in their university career mattered as well, with second-year students showing the highest rates of heavy drinking. These findings suggested that drinking too many soft drinks was not an isolated habit but often part of a broader cluster of lifestyle choices and specific beliefs about food and health.
To make sense of these connections and predict future behavior, the researchers fed all this data into seven different machine-learning algorithms. These were computer programs designed to learn from the data and classify students as either high or normal consumers. The team tested each program to see which one could make the most accurate predictions. The results showed that while all the models performed reasonably well, one stood out above the rest. A method called Random Forest, which works by combining the decisions of many simple decision trees to reach a more stable conclusion, proved to be the most effective. It correctly classified the students about 77 percent of the time and was particularly good at spotting the heavy drinkers, catching nearly 95 percent of them. This high success rate suggests that the combination of habits and attitudes the team measured contains enough information to reliably identify at-risk students.
Perhaps the most valuable insight from the study came from asking the winning computer model what it considered the most important clues. The algorithm identified two factors as the strongest predictors of high soft-drink consumption: smoking status and the student's academic year. This finding is significant because it highlights that the behavior is deeply intertwined with other lifestyle choices and the specific pressures or freedoms that come with different stages of university life. While the study was limited to a single university and relied on students' own reports, which can sometimes be imperfect, the results offer a clear direction for future action. The researchers suggest that health campaigns on campus should not just focus on the sugar content of the drinks but should also target the broader lifestyle patterns of students, particularly those who smoke or are in their second year of study. By understanding these specific drivers, universities can design more effective interventions to help young people make healthier beverage choices before these habits become entrenched for life.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.