Male student from the Faculty of Humanities develops AI system to predict university entrance exam scores in real time.

The idea for the system originated from Duc Anh's own experience of being overwhelmed by the sea of ​​admissions information. Initially, the student only intended to create a small tool to automate the filtering of admissions data from previous years. The turning point came when Duc Anh received guidance from Professor Tran Dang Ninh - a lecturer at the Academy of Security Technology. Under his in-depth guidance, the project transcended the boundaries of a typical tool to develop into a serious scientific research topic, combining big data processing and solving social problems.

Through his research, Duc Anh realized that university entrance exam scores are not just meaningless numbers on paper, but also directly reflect the supply and demand dynamics of the labor market and socio-psychological trends.

Nguyen Duc Anh, a student from the Information Management Faculty of the University of Social Sciences and Humanities, Vietnam National University, Hanoi, has developed an AI system that predicts university entrance exam scores in real time.

"If we only showed students last year's scores, that's 'dead' information. The real opportunity came when I started collecting and analyzing thousands of educational articles and realized the clear correlation between media attention and fluctuations in admission scores. From there, I decided to add machine learning models to the project, transforming a static search tool into a 'living' system capable of prediction," Duc Anh shared.

Based on a scientific research project, Duc Anh and his colleagues successfully developed the online platform diemthi.techtreesolution.com. This system not only supports searching and predicting cutoff scores, but also helps candidates assess risk and determine a safe score range for most universities nationwide.

Patiently sifting through raw data to find hidden treasures.
Duc Anh stated that instead of overusing complex but inefficient algorithms, the product's artificial intelligence system prioritizes stability and operates according to a three-step process: data collection, processing and storage, and analysis and prediction.

Automated code will periodically scan information from authoritative sources. The raw data is then cleaned and fed into a structured database. Finally, the massive dataset is run through natural language processing models to score the "popularity" of each field, combined with the 2026 exam score distribution to generate a safe predicted score range.

The real-time benchmark prediction system is used by tens of thousands of people.

"The biggest challenge is collecting and cleaning the data, because each school's admissions information has a different format. In the field of AI, the rule 'junk input data will produce junk results' is a significant challenge, especially with the diverse Vietnamese language and its many slang words," Duc Anh said.

To solve this problem, the student didn't rush to throw everything into the training model, but instead spent most of his time on the preprocessing step. Duc Anh built keyword filters, categorized news into clear topic clusters, and patiently trained the model with small, highly accurate datasets before scaling up.

Real-time forecasts help ease difficulties for test-takers.
"The most outstanding aspect of the project, as noted at the Student Scientific Research Conference, wasn't the sophisticated technology, but rather the real-time application of the concept to solve practical problems faced by students and parents," the male student shared.

Every 12 hours, algorithms automatically scan for the latest news and immediately process it to update the core database. Thanks to this data source being refreshed every half day, the predicted scores for a major will fluctuate continuously based on the actual trends in society. Candidates can always grasp the risk or safety of a major at the present moment, thus having the most solid basis for finalizing their list of preferences.

Sharing his future development plans, Duc Anh said: "For now, I'm still focusing on optimizing the 'heart' of the system, which is the search tool through filters, ensuring a smooth user experience. In the long term, I have two main goals. Firstly, to continue expanding the training dataset from social networks to make the prediction model more accurate. Secondly, I plan to dedicate my efforts to developing a mobile version."

According to the Vietnam Student Newspaper:Male student from the Faculty of Humanities develops AI system to predict university entrance exam scores in real time.