A series of csv files from several experiments of a group of 5 person is processed and converted into a final csv dataset ready to use in further AI-based modelling.
This repository shows Data preparation phase of Data Science course mini-project (SoSe 2022) at Saarland University.
Real time, raw data of Oxygen saturation (SpO2) and Heart Rate are taken from the sensor + patch developed by the Leibniz Institute of New Materials (INM).
Heart Rate and SpO2 values are measured while performing certain physiological activities such as “climbing steps”, “walking briskly”, “squats”, “planks” and “mountain climbing (MC)”.
Data Preparation consists on three steps:
-
Manual preprocessing:
- Construction of independent csv file containing SpO2 and Heart Rate for each experiment.
- Manual cleaning of junk data.
-
Processing exercise subsets:
- Remplacement of default sensor values “-999” by a mean value.
- Shrink data in order to have balanced information for each type of physical exercise.
-
Removing outliers:
- Outlier identification.
- Outlier removal: Standard Deviation Method.
- Outlier removal: DBSCAN.