Data Journal Machine Learning Jupyter Notebook

Machine Learning Project.

Submitted as part of a machine learning class in my pursuit of a Masters in Data Analytics through NW Missouri State University.

Github repo: https://github.com/aarongilly/ml_regression_gillespie

The most relevant part

What does this say about life? What does this say about machine learning?

The data features most highly correlated with satisfaction were surprising.

1. active_calories - from my Oura Ring, their documentation says this is a measure of “Active calories expended”
2. steps - also Oura Ring, “Total number of steps taken”, which would correlate highly with #1
3. satisfaction_running_avg - looking at the previous 3 days to make your prediction about today
4. work_status_endocded - weekends are generally more satisfying
5. health - my subjective health rating, 1 to 10, made at the same time I do the satisfaction rating

All of those were positively correlated with increased satisfaction. Interestingly, nothing in the top 5 has to do with sleep, actual workouts (as opposed to generic activity), and/or social outings.

The predictive power considering the strongest negative correlates were lower than the positive correlates. The most negatively correlated features were:

6. temperature_deviation - from my Oura Ring, “Temperature deviation in degrees Celsius.” - I believe during sleep, indicating some fluntuation in health and metabolism.
7. treatments - indicating some form of medication or physical therapy took place
8. pains - indicating something hurt enough to warrant tracking, almost 1:1 with treatments
9. lift_volume_lbs - one of the most surprising contraindications here, when I lift MORE, I feel less satisfied
10. awake_time - from my Oura Ring, time spent awake while trying to sleep the night before

[!tldr] If you want to be more satisfied: try to move more and work and be sick less…

…but there’s not much you can do to “hone” satisfaction levels using what’s tracked here.

You can track dozens of datapoints about your life, but that doesn’t make your subjective life satisfaction a highly-preditable phenomenon.

Machine Learning is an awesome technique. It can be used to reliably predict the weather many days from now using phenomenon observed today… but it’s not capable of devining results in a low-vailidity environment, such as the highly abstract nature of life satisfaction.