Projects
Prediction on Song Popularity Project
Using features of songs, such as energy, danceability, acousticness, key, loudness, tempo, etc. to predict the target variable: the popularity of songs, we built a prediction model to reach the goal of successfully predicting the popularity of a song. Machine learning and statistical models form the basis of our analysis.
For this project, we run the SQL function program in a miniature relational database with the order by using Python. Further, we optimize the required SQL function code with Python to observe the performance.
SMA Movie Analysis ML Models
Increased movie revenue forecasting accuracy by 30% after analysis of Kaggle and Twitter IMDb datasets by utilizing Python web-scraping to extract the top 1,000 movies and applying Ensemble Learning such as Bagging, Boosting, and AdaBoost to improve the final output.
Our team built a nutritional label for an ADS system that predicts heart disease. We also used mitigation techniques such as the generation of synthetic data, to create a model that can generalize better across subpopulations. In addition to fairness measures, we used interpretability tools including LIME and SHAP to observe the risk factors underlying a disease.