Lior Nattiv
Data Science Fellows February 2021 Cohort
Abstract
A regression/classification task with multiple target variables for time-series and tabular data based on the company’s remote pregnancy monitoring solution.
The project comprised of data preparation, data exploration, models building and performance evaluation with Various ML and Auto-ML libraries.
Challenges
- Doing very independent work.
- Getting to know and learn new Auto-ML libraries and how to work with them
Achievements
- Very good results with baseline and auto-ml models (Mostly sensitivity and specificity)
- Managed to test many different models during the process
- Managed to take one of the top models found via auto-ml process and improve results by using the same model from Sk-learn’s algorithms
Future project development
Please describe possible further development of the project following your project phase.
- More sample data will increase reliability of results, and will be able to help build much more efficient regression models (which was not possible with the current data of the project)
- Building a Deep-Learning neural network with raw data. It was planned, but there was not enough time to do so.
- Testing additional Auto-ML libraries.