Yonsei University, 2019. 3. ~ 2019. 7.
Project Summary
It was conducted as a graduation project of Yonsei university computer science department. The project is about applying a recommendation algorithm to infer correlations between specific areas of the brain and keywords(emotions, illnesses, activities, etc). We trained a collaborative filtering model with the dataset and visualized the results.
A keyword is given as a query(angry above). Then the red ones are an brain area known to be correlated with the query. The blue ones are the brain area predicted to be correlated by the recommendation algorithm. Users can view the ineference results for the 1335 keywords.
We validated our models in two ways, k-fold validation and time-series validation. We used AUROC(area under ROC curve) as a metric. We calculated AUROC of each keyward and drawed histograms. It showed that our model gave some reliable results.
Role
- Lead the team as a leader
- Data engineering of the raw brain voxel data
- Researched and implemented recommendation algorithms
- Visualized inference results
Tech Stack
- Language: python
- Library: numpy, pandas
Results
- Best Award of 2019 Spring Yonsei Univ Software Capstone Project