AUTOML REGRESSION MODEL TO PREDICT VIEWING AND LISTENING BEHAVIORS

Created a scalable Auto Machine Learning - Regression pipeline that predicts user behavior (n° of minutes of video content watched) in 4 countries. The AutoML pipeline fits 50+ different Machine Learning algorithms, compares their performance, selects the best-performing one, fine-tunes it, visualizes various interactive charts, and makes the final predictions. The model reached an R2 > 80%, leading to better forecasts and consequent better planning of more targeted ads in most watched part of the day by users of interest.