Using NumPy to Analyze My Daily Habits (Sleep, Screen Time & Mood)

a small NumPy venture collection the place I attempt to truly construct one thing with NumPy as an alternative of simply going by random features and documentation. I’ve at all times felt that one of the simplest ways to study is by doing, so on this venture, I needed to create one thing each sensible and private.

The concept was easy: analyze my day by day habits — sleep, research hours, display time, train, and temper — and see how they have an effect on my productiveness and common well-being. The info isn’t actual; it’s fictional, simulated over 30 days. However the aim isn’t the accuracy of the info — it’s studying tips on how to use NumPy meaningfully.

So let’s stroll by the method step-by-step.

Step 1 — Loading and Understanding the Knowledge

I began by making a easy NumPy array that contained 30 rows (one for every day) and 6 columns — every column representing a unique behavior metric. Then I saved it as a .npy file so I may simply load it later.

# TODO: Import NumPy and cargo the .npy information file
import numpy as np
information = np.load(‘activity_data.npy’)

As soon as loaded, I needed to verify that every part seemed as anticipated. So I checked the form (to know what number of rows and columns there have been) and the variety of dimensions (to verify it’s a 2D desk, not a 1D listing).

# TODO: Print array form, first few rows, and so forth.
information.form
information.ndim

OUTPUT: 30 rows, 6 columns, and ndim=2

I additionally printed out the primary few rows simply to visually affirm that every worth seemed nice — for example, that sleep hours weren’t adverse or that the temper values have been inside an inexpensive vary.

# TODO: Prime 5 rows
information[:5]

Output:

array([[ 1. , 6.5, 5. , 4.2, 20. , 6. ],
[ 2. , 7.2, 6. , 3.1, 35. , 7. ],
[ 3. , 5.8, 4. , 5.5, 0. , 5. ],
[ 4. , 8. , 7. , 2.5, 30. , 8. ],
[ 5. , 6. , 5. , 4.8, 10. , 6. ]])

Step 2 — Validating the Knowledge

Earlier than doing any evaluation, I needed to ensure the info made sense. It’s one thing we frequently skip when working with fictional information, nevertheless it’s nonetheless good apply.

So I checked:

• No adverse sleep hours
• No temper scores lower than 1 or better than 10

For sleep, that meant deciding on the sleep column (index 1 in my array) and checking if any values have been beneath zero.

# Make sure that values are cheap (no adverse sleep)
information[:, 1] < 0

Output:

array([False, False, False, False, False, False, False, False, False,
False, False, False, False, False, False, False, False, False,
False, False, False, False, False, False, False, False, False,
False, False, False])

This implies no negatives. Then I did the identical for temper. I counted to search out that the temper column was at index 5, and checked if any have been beneath 1 or above 10.

# Is temper out of vary?
information[:, 5] < 1
information[:, 5] > 10

We received the identical output.

The whole lot seemed good, so we may transfer on.

Step 3 — Splitting the Knowledge into Weeks

I had 30 days of information, and I needed to investigate it week by week. The primary intuition was to make use of NumPy’s break up() perform, however that failed as a result of 30 isn’t evenly divisible by 4. So as an alternative, I used np.array_split(), which permits uneven splits.

That gave me:

• Week 1 → 8 days
• Week 2 → 8 days
• Week 3 → 7 days
• Week 4 → 7 days

# TODO: Slice information into week 1, week 2, week 3, week 4
weekly_data = np.array_split(information, 4)
weekly_data

Output:

[array([[ 1. , 6.5, 5. , 4.2, 20. , 6. ],
[ 2. , 7.2, 6. , 3.1, 35. , 7. ],
[ 3. , 5.8, 4. , 5.5, 0. , 5. ],
[ 4. , 8. , 7. , 2.5, 30. , 8. ],
[ 5. , 6. , 5. , 4.8, 10. , 6. ],
[ 6. , 7.5, 6. , 3.3, 25. , 7. ],
[ 7. , 8.2, 3. , 6.1, 40. , 7. ],
[ 8. , 6.3, 4. , 5. , 15. , 6. ]]),

array([[ 9. , 7. , 6. , 3.2, 30. , 7. ],
[10. , 5.5, 3. , 6.8, 0. , 5. ],
[11. , 7.8, 7. , 2.9, 25. , 8. ],
[12. , 6.1, 5. , 4.5, 15. , 6. ],
[13. , 7.4, 6. , 3.7, 30. , 7. ],
[14. , 8.1, 2. , 6.5, 50. , 7. ],
[15. , 6.6, 5. , 4.1, 20. , 6. ],
[16. , 7.3, 6. , 3.4, 35. , 7. ]]),

array([[17. , 5.9, 4. , 5.6, 5. , 5. ],
[18. , 8.3, 7. , 2.6, 30. , 8. ],
[19. , 6.2, 5. , 4.3, 10. , 6. ],
[20. , 7.6, 6. , 3.1, 25. , 7. ],
[21. , 8.4, 3. , 6.3, 40. , 7. ],
[22. , 6.4, 4. , 5.1, 15. , 6. ],
[23. , 7.1, 6. , 3.3, 30. , 7. ]]),

array([[24. , 5.7, 3. , 6.7, 0. , 5. ],
[25. , 7.9, 7. , 2.8, 25. , 8. ],
[26. , 6.2, 5. , 4.4, 15. , 6. ],
[27. , 7.5, 6. , 3.5, 30. , 7. ],
[28. , 8. , 2. , 6.4, 50. , 7. ],
[29. , 6.5, 5. , 4.2, 20. , 6. ],
[30. , 7.4, 6. , 3.6, 35. , 7. ]])]

Now the info was in 4 chunks, and I may simply analyze each individually.

Step 4 — Calculating Weekly Metrics

I needed to get a way of how every behavior modified from week to week. So I targeted on 4 major issues:

• Common sleep
• Common research hours
• Common display time
• Common temper rating

I saved every week’s array in a separate variable, then used np.imply() to calculate the averages for every metric.

Common sleep hours

# retailer into variables
week_1 = weekly_data[0]
week_2 = weekly_data[1]
week_3 = weekly_data[2]
week_4 = weekly_data[3]

# TODO: Compute common sleep
week1_avg_sleep = np.imply(week_1[:, 1])
week2_avg_sleep = np.imply(week_2[:, 1])
week3_avg_sleep = np.imply(week_3[:, 1])
week4_avg_sleep = np.imply(week_4[:, 1])

Common research hours

# TODO: Compute common research hours
week1_avg_study = np.imply(week_1[:, 2])
week2_avg_study = np.imply(week_2[:, 2])
week3_avg_study = np.imply(week_3[:, 2])
week4_avg_study = np.imply(week_4[:, 2])

Common display time

# TODO: Compute common display time
week1_avg_screen = np.imply(week_1[:, 3])
week2_avg_screen = np.imply(week_2[:, 3])
week3_avg_screen = np.imply(week_3[:, 3])
week4_avg_screen = np.imply(week_4[:, 3])

Common temper rating

# TODO: Compute common temper rating
week1_avg_mood = np.imply(week_1[:, 5])
week2_avg_mood = np.imply(week_2[:, 5])
week3_avg_mood = np.imply(week_3[:, 5])
week4_avg_mood = np.imply(week_4[:, 5])

Then, to make every part simpler to learn, I formatted the outcomes properly.

# TODO: Show weekly outcomes clearly
print(f”Week 1 — Common sleep: {week1_avg_sleep:.2f} hrs, Examine: {week1_avg_study:.2f} hrs, “
f”Display screen time: {week1_avg_screen:.2f} hrs, Temper rating: {week1_avg_mood:.2f}”)

print(f”Week 2 — Common sleep: {week2_avg_sleep:.2f} hrs, Examine: {week2_avg_study:.2f} hrs, “
f”Display screen time: {week2_avg_screen:.2f} hrs, Temper rating: {week2_avg_mood:.2f}”)

print(f”Week 3 — Common sleep: {week3_avg_sleep:.2f} hrs, Examine: {week3_avg_study:.2f} hrs, “
f”Display screen time: {week3_avg_screen:.2f} hrs, Temper rating: {week3_avg_mood:.2f}”)

print(f”Week 4 — Common sleep: {week4_avg_sleep:.2f} hrs, Examine: {week4_avg_study:.2f} hrs, “
f”Display screen time: {week4_avg_screen:.2f} hrs, Temper rating: {week4_avg_mood:.2f}”)

Output:

Week 1 – Common sleep: 6.94 hrs, Examine: 5.00 hrs, Display screen time: 4.31 hrs, Temper rating: 6.50
Week 2 – Common sleep: 6.97 hrs, Examine: 5.00 hrs, Display screen time: 4.39 hrs, Temper rating: 6.62
Week 3 – Common sleep: 7.13 hrs, Examine: 5.00 hrs, Display screen time: 4.33 hrs, Temper rating: 6.57
Week 4 – Common sleep: 7.03 hrs, Examine: 4.86 hrs, Display screen time: 4.51 hrs, Temper rating: 6.57

Step 5 — Making Sense of the Outcomes

As soon as I printed out the numbers, some patterns began to indicate up.

My sleep hours have been fairly regular for the primary two weeks (round 6.9 hours), however in week three, they jumped to round 7.1 hours. Meaning I used to be “sleeping higher” because the month went on. By week 4, it stayed roughly round 7.0 hours.

For research hours, it was the other. Week one and two had a mean of round 5 hours per day, however by week 4, it had dropped to about 4 hours. Principally, I began off sturdy however slowly misplaced momentum — which, actually, sounds about proper.

Then got here display time. This one damage a bit. In week one, it was roughly 4.3 hours per day, and it simply saved creeping up each week. The traditional cycle of being productive early on, then slowly drifting into extra “scrolling breaks” later within the month.

Lastly, there was temper. My temper rating began at round 6.5 in week one, went barely as much as 6.6 in week two, after which type of hovered there for the remainder of the interval. It didn’t transfer dramatically, nevertheless it was fascinating to see a small spike in week two — proper earlier than my research hours dropped and my display time elevated.

To make issues interactive, I believed it’d be nice to visualise utilizing matplotlib.

Step 6 — On the lookout for Patterns

Now that I had the numbers, I needed to know why my temper went up in week two.

So I in contrast the weeks aspect by aspect. Week two had respectable sleep, excessive research hours, and comparatively low display time in comparison with the later weeks.

That may clarify why my temper rating peaked there. By week three, regardless that I slept extra, my research hours had began to dip — perhaps I used to be resting extra however getting much less completed, which didn’t increase my temper as a lot as I anticipated.

That is what I preferred concerning the venture: it’s not concerning the information being actual, however about how one can use NumPy to discover patterns, relationships, and small insights. Even fictional information can inform a narrative whenever you take a look at it the suitable means.

Step 7 — Wrapping Up and Subsequent Steps

On this little venture, I realized a couple of key issues — each about NumPy and about structuring evaluation like this.

We began with a uncooked array of fictional day by day habits, realized tips on how to test its construction and validity, break up it into significant chunks (weeks), after which used easy NumPy operations to investigate every phase.

It’s the type of small venture that reminds you that information evaluation doesn’t at all times must be complicated. Typically it’s nearly asking easy questions like “How is my display time altering over time?” or “When do I really feel the perfect?”

If I needed to take this additional (which I in all probability will), there are such a lot of instructions to go:

• Discover the greatest and worst days total
• Examine weekdays vs weekends
• And even create a easy “wellbeing rating” primarily based on a number of habits mixed

However that’ll in all probability be for the subsequent a part of the collection.

For now, I’m glad that I received to use NumPy to one thing that feels actual and relatable — not simply summary arrays and numbers, however habits and feelings. That’s the type of studying that sticks.

Thanks for studying.

For those who’re following together with the collection, strive recreating this by yourself fictional information. Even when your numbers are random, the method will educate you tips on how to slice, break up, and analyze arrays like a professional.