Analyzed my personal sleep data from my Fitbit Charge 5 in combination with the lunar phases.
Moon Phases
Date: 03.04.2022
Source: https://www.timeanddate.de/mond/phasen/ (Location: Basel):
raw/moon/*
Sleep Data
Date: 18.06.2022 Source: export of complete fitbit data of privat account files:
raw/sleep/*raw/heart/*raw/temp/*
- Cleaning and transformation of data
- Merge data sets on attribute 'date'
Snippet of the Master-DataFrame after wrangling:
| date | skin_temp_avg | overall_score | composition_score | revitalization_score | duration_score | deep_sleep_in_minutes | resting_heart_rate | restlessness | avg_bpm | Moon Phase |
|---|---|---|---|---|---|---|---|---|---|---|
| 04.03.22 | 30.505155 | 67 | 17 | 15 | 35 | 66 | 60 | 0.073479 | 66.494309 | New Moon |
| 19.04.22 | 30.625221 | 75 | 16 | 17 | 42 | 62 | 52 | 0.071966 | 58.762171 | |
| 10.03.22 | 29.899359 | 88 | 21 | 22 | 45 | 126 | 60 | 0.054695 | 60.620142 | First Quarter |
| 25.04.22 | 29.917705 | 56 | 15 | 15 | 26 | 23 | 56 | 0.100186 | 60.969515 | Third Quarter |
| 08.06.22 | 30.016326 | 75 | 20 | 19 | 36 | 69 | 55 | 0.087816 | 59.398694 | First Quarter |
| 01.01.22 | 30.85856 | 68 | 19 | 18 | 31 | 69 | 60 | 0.081197 | 58.885041 | New Moon |
| 15.01.22 | 32.980144 | 61 | 17 | 10 | 34 | 70 | 57 | 0.14459 | 66.889982 | |
| 14.01.22 | 31.024239 | 77 | 21 | 20 | 36 | 82 | 59 | 0.089701 | 64.67045 | |
| 09.06.22 | 30.251775 | 76 | 22 | 16 | 38 | 88 | 54 | 0.07064 | 53.8481 | First Quarter |
| 24.04.22 | 30.351842 | 68 | 16 | 15 | 37 | 49 | 54 | 0.079038 | 68.521044 | Third Quarter |
I looked at the distributions etc. Furthermore, the moon phases do not seem to have a significant influence on my sleep:
-> some specific preprocessing for the algorithms (one hot encoding, shortening of column names, etc ...)
Sklearn Validation Set Measured vs Fitted
MSE: 5.1
Sklearn Test Set Measured vs Fitted
Test Set MSE: 6.9
MSE: 4.9
