title: "Fitbit Analysis"

author: "Brandi Falls"

date: "7/10/2021"

---

Overview of Project

Bellabeat is a high-tech manufacturer of health-focused products for women. Bellabeat is a successful small company, but they have the potential to become a larger player in the global smart device. Analyzing smart device fitness data could help unlock new growth opportunities for the company market.

---

Step 1 - Ask

Business Task

The business task is to analyze Fitbit smart device usage data in order to gain insight into how consumers currently use their smart devices and to discover underlying trends that can be applied to Bellabeat products. This insight could help unlock new growth opportunities and the best marketing strategy to meet business growth goals.

---

Step 2 - Prepare

Description of the Dataset

We have used a public dataset that explores smart device users’ daily habits.

Name: FitBit Fitness Tracker Data (CC0 Public Domain) Link: https://www.kaggle.com/arashnic/fitbit

"This Kaggle data set contains personal fitness tracker from thirty fitbit users. Thirty eligible Fitbit users consented to the submission of personal tracker data, including minute-level output for physical activity, heart rate, and sleep monitoring. It includes information about daily activity, steps, and heart rate that can be used to explore users’ habits.

These datasets were generated by respondents to a distributed survey via Amazon Mechanical Turk between 03.12.2016-05.12.2016. Thirty eligible Fitbit users consented to the submission of personal tracker data, including minute-level output for physical activity, heart rate, and sleep monitoring. "

Limitations

• There are limited users. The sample size may not be representative of the population.
• There are only a few categorical variables to categorize the data for analysis.
• There is no data on male and female population. We will not be able to make female-specific observations for the female target audience.
• Each dataset has 30 days of data from 4/12/2016 to 5/12/2016, but the dataset description indicates 03/12/2016 to 05/12/2016. We will have less data to work with which may impact recommendations.
• The first day of data collected is Tuesday 4/12/2016, and the last day of data collected is on Thursday 5/12/2016. There is an unbalanced spread of days, so any weekly analysis could be misleading.

ROCCC Validation of Dataset

• Reliable - We will assume this data is accurate and reliable.
• Original - The data is original as it is directly gathered from users
• Comprehensive - The data is not comprehensive as it does not have all details needed for a full detailed analysis to meet the business goal
• Current - The data is not current as the data is from 2016.
• Cited - The data is cited.

Data Summary

Overall, this data may not be appropriate for conducting a full detailed analysis to meet the goal of the business. However, we will move forward with the data to uncover any insights.

Ideally, we would be having a sample size that is more representative of the whole population, at least 6 months of data from a more current year, and gendered data so that we can gain insight into the specific female trends.

Data Sensitivity

This is public data. No data was identified as sensitive and therefore we will not need to have Data anonymization / De-identification of any data. Also, there is no need to encrypt the data.

Setup Environment with Packages

#import libraries
import pandas as pd
import numpy as np
import seaborn as sns
import janitor
import datetime

import matplotlib.pyplot as plt
import matplotlib.mlab as mlab
import matplotlib
from matplotlib.pyplot import figure
from IPython.core.interactiveshell import InteractiveShell
InteractiveShell.ast_node_interactivity = "all"

%matplotlib inline
matplotlib.rcParams['figure.figsize'] = (12,8)

pd.options.mode.chained_assignment = None

Import Data

• The below datasets were determined to be most useful for the purpose of this analysis. All other available datasets will not be in scope.
  • dailyActivity (daily physical activity)
  • sleepDay (daily sleep activity)
  • weightLogInfo (daily weight log activity)

#load data from csv files into data frames
dailyActivity = pd.read_csv(r'C:/Users/Brandi/Documents/Bellabeat Case Study/fitbit/Fitabase Data 4.12.16-5.12.16/dailyActivity_merged.csv')
sleepDay = pd.read_csv(r'C:/Users/Brandi/Documents/Bellabeat Case Study/fitbit/Fitabase Data 4.12.16-5.12.16/sleepDay_merged.csv')
weightLogInfo = pd.read_csv(r'C:/Users/Brandi/Documents/Bellabeat Case Study/fitbit/Fitabase Data 4.12.16-5.12.16/weightLogInfo_merged.csv')

First Look

dailyActivity.head()
sleepDay.head()
weightLogInfo.head()

	Id	ActivityDate	TotalSteps	TotalDistance	TrackerDistance	LoggedActivitiesDistance	VeryActiveDistance	ModeratelyActiveDistance	LightActiveDistance	SedentaryActiveDistance	VeryActiveMinutes	FairlyActiveMinutes	LightlyActiveMinutes	SedentaryMinutes	Calories
0	1503960366	4/12/2016	13162	8.50	8.50	0.0	1.88	0.55	6.06	0.0	25	13	328	728	1985
1	1503960366	4/13/2016	10735	6.97	6.97	0.0	1.57	0.69	4.71	0.0	21	19	217	776	1797
2	1503960366	4/14/2016	10460	6.74	6.74	0.0	2.44	0.40	3.91	0.0	30	11	181	1218	1776
3	1503960366	4/15/2016	9762	6.28	6.28	0.0	2.14	1.26	2.83	0.0	29	34	209	726	1745
4	1503960366	4/16/2016	12669	8.16	8.16	0.0	2.71	0.41	5.04	0.0	36	10	221	773	1863

	Id	SleepDay	TotalSleepRecords	TotalMinutesAsleep	TotalTimeInBed
0	1503960366	4/12/2016 12:00:00 AM	1	327	346
1	1503960366	4/13/2016 12:00:00 AM	2	384	407
2	1503960366	4/15/2016 12:00:00 AM	1	412	442
3	1503960366	4/16/2016 12:00:00 AM	2	340	367
4	1503960366	4/17/2016 12:00:00 AM	1	700	712

	Id	Date	WeightKg	WeightPounds	Fat	BMI	IsManualReport	LogId
0	1503960366	5/2/2016 11:59:59 PM	52.599998	115.963147	22.0	22.650000	True	1462233599000
1	1503960366	5/3/2016 11:59:59 PM	52.599998	115.963147	NaN	22.650000	True	1462319999000
2	1927972279	4/13/2016 1:08:52 AM	133.500000	294.317120	NaN	47.540001	False	1460509732000
3	2873212765	4/21/2016 11:59:59 PM	56.700001	125.002104	NaN	21.450001	True	1461283199000
4	2873212765	5/12/2016 11:59:59 PM	57.299999	126.324875	NaN	21.690001	True	1463097599000

Observations from First Look

• Column names need to be cleaned
• Common variables are ID and ActivityDate/ActivityDay.
• Date variables are in format MM/DD/YYYY but are Character data type.
• We will need to rename Day --> Date for consistency, and convert to Date field.
• LoggedActivitiesDistance has a lot of 0 values. Assumption is that it is not required to log most activities, but some activities require manual entry. Users may rely on automated tracking and will not manually log activities.
• The file sleepDay should be renamed to dailySleep, and SleepDay variable should be renamed to ActivityDate and formatted to MM/DD/YYYY for consistency

---

Step 3 - Process

Summary of Cleaning:

1. Created new dataframes with simple names for easy reference
2. Cleaned variable names (changed from PascalCase to snake_case)
3. Checked for missing (NA/null) data
4. Checked for duplicate data
5. Stored and then removed duplicate data
6. Renamed variables for consistency and joining
7. Converted all Date variables from Character to Date
8. Organized data by sorting by date
9. Converted ID from Number to Character for display purposes
10. Added a Day column to categorize data by days
11. Added a Record Type variable for categorical purposes
12. Performed a Set Difference between additional data to confirm all data was already present in the merged dataset DailyActivity

#checking for missing data (NULL)
for col in dailyActivity.columns:
    pct_missing = round((np.mean(dailyActivity[col].isnull())) * 100)
    print(f'{col} {pct_missing}%')

print("---")

for col in sleepDay.columns:
    pct_missing = round((np.mean(sleepDay[col].isnull())) * 100)
    print(f'{col} {pct_missing}%')

print("---")

for col in weightLogInfo.columns:
    pct_missing = round((np.mean(weightLogInfo[col].isnull())) * 100)
    print(f'{col} {pct_missing}%')

#checking for missing data (NA)
for col in dailyActivity.columns:
    pct_missing = round((np.mean(dailyActivity[col].isna())) * 100)
    print(f'{col} {pct_missing}%')

print("---")

for col in sleepDay.columns:
    pct_missing = round((np.mean(sleepDay[col].isna())) * 100)
    print(f'{col} {pct_missing}%')

print("---")

for col in weightLogInfo.columns:
    pct_missing = round((np.mean(weightLogInfo[col].isna())) * 100)
    print(f'{col} {pct_missing}%')

# weightLogInfo["Fat"] has missing data, but we will retain it. 

Id 0%
ActivityDate 0%
TotalSteps 0%
TotalDistance 0%
TrackerDistance 0%
LoggedActivitiesDistance 0%
VeryActiveDistance 0%
ModeratelyActiveDistance 0%
LightActiveDistance 0%
SedentaryActiveDistance 0%
VeryActiveMinutes 0%
FairlyActiveMinutes 0%
LightlyActiveMinutes 0%
SedentaryMinutes 0%
Calories 0%
---
Id 0%
ActivityDate 0%
TotalSteps 0%
TotalDistance 0%
TrackerDistance 0%
LoggedActivitiesDistance 0%
VeryActiveDistance 0%
ModeratelyActiveDistance 0%
LightActiveDistance 0%
SedentaryActiveDistance 0%
VeryActiveMinutes 0%
FairlyActiveMinutes 0%
LightlyActiveMinutes 0%
SedentaryMinutes 0%
Calories 0%
---
Id 0%
SleepDay 0%
TotalSleepRecords 0%
TotalMinutesAsleep 0%
TotalTimeInBed 0%
---
Id 0%
Date 0%
WeightKg 0%
WeightPounds 0%
Fat 97%
BMI 0%
IsManualReport 0%
LogId 0%
Id 0%
ActivityDate 0%
TotalSteps 0%
TotalDistance 0%
TrackerDistance 0%
LoggedActivitiesDistance 0%
VeryActiveDistance 0%
ModeratelyActiveDistance 0%
LightActiveDistance 0%
SedentaryActiveDistance 0%
VeryActiveMinutes 0%
FairlyActiveMinutes 0%
LightlyActiveMinutes 0%
SedentaryMinutes 0%
Calories 0%
---
Id 0%
SleepDay 0%
TotalSleepRecords 0%
TotalMinutesAsleep 0%
TotalTimeInBed 0%
---
Id 0%
Date 0%
WeightKg 0%
WeightPounds 0%
Fat 97%
BMI 0%
IsManualReport 0%
LogId 0%

# Checking data types
dailyActivity.dtypes
sleepDay.dtypes
weightLogInfo.dtypes

Id                            int64
ActivityDate                 object
TotalSteps                    int64
TotalDistance               float64
TrackerDistance             float64
LoggedActivitiesDistance    float64
VeryActiveDistance          float64
ModeratelyActiveDistance    float64
LightActiveDistance         float64
SedentaryActiveDistance     float64
VeryActiveMinutes             int64
FairlyActiveMinutes           int64
LightlyActiveMinutes          int64
SedentaryMinutes              int64
Calories                      int64
dtype: object

Id                     int64
SleepDay              object
TotalSleepRecords      int64
TotalMinutesAsleep     int64
TotalTimeInBed         int64
dtype: object

Id                  int64
Date               object
WeightKg          float64
WeightPounds      float64
Fat               float64
BMI               float64
IsManualReport       bool
LogId               int64
dtype: object

#cleaning data frames
#changing dataframe names, standardize col names, remove empty rows, renaming columns
updated_daily_activity = (
    dailyActivity
      .clean_names(None, 'snake')
      .remove_empty()
)

updated_daily_sleep = (
    sleepDay
      .clean_names(None, 'snake')
      .remove_empty()
      .rename_column('sleep_day','activity_date')
)

updated_weight_log = (
    weightLogInfo
      .clean_names(None, 'snake')
      .remove_empty()
      .rename_column('date',"activity_date")
)

#changing activity date string to datetime 
updated_daily_activity['activity_date'] = pd.to_datetime(updated_daily_activity['activity_date'])
updated_daily_sleep['activity_date'] = pd.to_datetime(updated_daily_sleep['activity_date'])
updated_weight_log['activity_date'] = pd.to_datetime(updated_weight_log['activity_date'])

#adding weekday column
updated_daily_activity['weekday'] = updated_daily_activity['activity_date'].dt.day_name()
updated_daily_sleep['weekday'] = updated_daily_sleep['activity_date'].dt.day_name()
updated_weight_log['weekday'] = updated_weight_log['activity_date'].dt.day_name()

#change data types of ID from int to string
updated_daily_activity['id'] = updated_daily_activity['id'].astype(str)
updated_daily_sleep['id'] = updated_daily_sleep['id'].astype(str)
updated_weight_log['id'] = updated_weight_log['id'].astype(str)

#check for duplicate data
updated_daily_activity.duplicated().sum()
updated_daily_sleep.duplicated().sum()
updated_weight_log.duplicated().sum()

0

3

0

updated_daily_sleep.loc[updated_daily_sleep.duplicated(), :]

	id	activity_date	total_sleep_records	total_minutes_asleep	total_time_in_bed	weekday
161	4388161847	2016-05-05	1	471	495	Thursday
223	4702921684	2016-05-07	1	520	543	Saturday
380	8378563200	2016-04-25	1	388	402	Monday

updated_daily_sleep = updated_daily_sleep.drop_duplicates()
updated_daily_sleep.duplicated().sum()

0

# sort by Date
updated_daily_activity = updated_daily_activity.sort_values(by=['activity_date'], inplace=False, ascending=True)
updated_daily_sleep = updated_daily_sleep.sort_values(by=['activity_date'], inplace=False, ascending=True)
updated_weight_log = updated_weight_log.sort_values(by=['activity_date'], inplace=False, ascending=True)

#add report type variable for categorical data
updated_weight_log['report_type'] = np.where(updated_weight_log['is_manual_report'] == True, "Manual", "Automated")

#validating new reoport type column
np.sum(updated_weight_log["report_type"] == "Manual")
np.sum(updated_weight_log["is_manual_report"] == True)

np.sum(updated_weight_log["report_type"] == "Automated")
np.sum(updated_weight_log["is_manual_report"] == False)

41

41

26

26

#comapre data frames
calorie_comparison_test = updated_daily_activity[["id","activity_date","calories"]]

intensity_comparison_test = updated_daily_activity[["id","activity_date","sedentary_minutes","lightly_active_minutes", "fairly_active_minutes","very_active_minutes","sedentary_active_distance","light_active_distance","moderately_active_distance","very_active_distance"]]

daily_steps_comparison_test = updated_daily_activity[["id","activity_date","total_steps"]]

dailyCalories = pd.read_csv(r'C:/Users/Brandi/Documents/Bellabeat Case Study/fitbit/Fitabase Data 4.12.16-5.12.16/dailyCalories_merged.csv')
dailyIntensities  = pd.read_csv(r'C:/Users/Brandi/Documents/Bellabeat Case Study/fitbit/Fitabase Data 4.12.16-5.12.16/dailyIntensities_merged.csv')
dailySteps  = pd.read_csv(r'C:/Users/Brandi/Documents/Bellabeat Case Study/fitbit/Fitabase Data 4.12.16-5.12.16/dailySteps_merged.csv')

a1 = (
    dailyCalories
      .clean_names(None, 'snake')
      .remove_empty()
      .rename_column('activity_day','activity_date')
)

a2 = (
    dailyIntensities
      .clean_names(None, 'snake')
      .remove_empty()
      .rename_column('activity_day','activity_date')
)

a3 = (
    dailySteps
      .clean_names(None, 'snake')
      .remove_empty()
      .rename_column('activity_day','activity_date')
      .rename_column('step_total','total_steps')
)

a1['id'] = a1['id'].astype(str)
a2['id'] = a2['id'].astype(str)
a3['id'] = a3['id'].astype(str)

a1['activity_date'] = pd.to_datetime(a1['activity_date'])
a2['activity_date'] = pd.to_datetime(a2['activity_date'])
a3['activity_date'] = pd.to_datetime(a3['activity_date'])

#looking for differences between data sets, none. 
pd.concat([calorie_comparison_test,a1]).drop_duplicates(keep=False)
pd.concat([intensity_comparison_test,a2]).drop_duplicates(keep=False)
pd.concat([daily_steps_comparison_test,a3]).drop_duplicates(keep=False)

	id	activity_date	calories

	id	activity_date	sedentary_minutes	lightly_active_minutes	fairly_active_minutes	very_active_minutes	sedentary_active_distance	light_active_distance	moderately_active_distance	very_active_distance

	id	activity_date	total_steps

---

Step 4 - Analyze

Summary of Analysis Process

1. Counted distinct participants in each dataframe
2. Examined date ranges in each dataframe
3. Merged dataframes
4. Created an order for days of the week for visualizations
5. Viewed summary statistics of each dataframe
6. Analyzed percentages of manual vs automated activity logs
7. Analyzed percentages of participants of each category

#quick view of clean data
updated_daily_activity.head()
updated_daily_sleep.head()
updated_weight_log.head()

	id	activity_date	total_steps	total_distance	tracker_distance	logged_activities_distance	very_active_distance	moderately_active_distance	light_active_distance	sedentary_active_distance	very_active_minutes	fairly_active_minutes	lightly_active_minutes	sedentary_minutes	calories	weekday
0	1503960366	2016-04-12	13162	8.50	8.50	0.0	1.88	0.55	6.06	0.0	25	13	328	728	1985	Tuesday
123	1927972279	2016-04-12	678	0.47	0.47	0.0	0.00	0.00	0.47	0.0	0	0	55	734	2220	Tuesday
154	2022484408	2016-04-12	11875	8.34	8.34	0.0	3.31	0.77	4.26	0.0	42	14	227	1157	2390	Tuesday
909	8877689391	2016-04-12	23186	20.40	20.40	0.0	12.22	0.34	7.82	0.0	85	7	312	1036	3921	Tuesday
185	2026352035	2016-04-12	4414	2.74	2.74	0.0	0.19	0.35	2.20	0.0	3	8	181	706	1459	Tuesday

	id	activity_date	total_sleep_records	total_minutes_asleep	total_time_in_bed	weekday
0	1503960366	2016-04-12	1	327	346	Tuesday
109	4020332650	2016-04-12	1	501	541	Tuesday
167	4445114986	2016-04-12	2	429	457	Tuesday
200	4702921684	2016-04-12	1	425	439	Tuesday
228	5553957443	2016-04-12	1	441	464	Tuesday

	id	activity_date	weight_kg	weight_pounds	fat	bmi	is_manual_report	log_id	weekday	report_type
43	8877689391	2016-04-12 06:47:11	85.800003	189.156628	NaN	25.680000	False	1460443631000	Tuesday	Automated
13	6962181067	2016-04-12 23:59:59	62.500000	137.788914	NaN	24.389999	True	1460505599000	Tuesday	Manual
2	1927972279	2016-04-13 01:08:52	133.500000	294.317120	NaN	47.540001	False	1460509732000	Wednesday	Automated
44	8877689391	2016-04-13 06:55:00	84.900002	187.172464	NaN	25.410000	False	1460530500000	Wednesday	Automated
14	6962181067	2016-04-13 23:59:59	62.099998	136.907061	NaN	24.240000	True	1460591999000	Wednesday	Manual

#unique participants in each dataframe
updated_daily_activity["id"].nunique()
updated_daily_sleep["id"].nunique()
updated_weight_log["id"].nunique()

33

24

8

#Examining Date Ranges. All date ranges are the same
updated_daily_activity["activity_date"].min()
updated_daily_sleep["activity_date"].min()
updated_weight_log["activity_date"].min()

print("---")

updated_daily_activity["activity_date"].max()
updated_daily_sleep["activity_date"].max()
updated_weight_log["activity_date"].max()

print("---")

updated_daily_activity["activity_date"].max() - updated_daily_activity["activity_date"].min()
updated_daily_sleep["activity_date"].max() - updated_daily_sleep["activity_date"].min()
updated_weight_log["activity_date"].max() - updated_weight_log["activity_date"].min()

Timestamp('2016-04-12 00:00:00')

Timestamp('2016-04-12 00:00:00')

Timestamp('2016-04-12 06:47:11')

---

Timestamp('2016-05-12 00:00:00')

Timestamp('2016-05-12 00:00:00')

Timestamp('2016-05-12 23:59:59')

---

Timedelta('30 days 00:00:00')

Timedelta('30 days 00:00:00')

Timedelta('30 days 17:12:48')

#quick summary statistics
updated_daily_activity[["total_steps", "total_distance", "sedentary_minutes"]].describe()
updated_daily_sleep [["total_sleep_records", "total_minutes_asleep", "total_time_in_bed"]].describe()
updated_weight_log [["weight_pounds", "bmi", "fat"]].describe()

	total_steps	total_distance	sedentary_minutes
count	940.000000	940.000000	940.000000
mean	7637.910638	5.489702	991.210638
std	5087.150742	3.924606	301.267437
min	0.000000	0.000000	0.000000
25%	3789.750000	2.620000	729.750000
50%	7405.500000	5.245000	1057.500000
75%	10727.000000	7.712500	1229.500000
max	36019.000000	28.030001	1440.000000

	total_sleep_records	total_minutes_asleep	total_time_in_bed
count	410.000000	410.000000	410.000000
mean	1.119512	419.173171	458.482927
std	0.346636	118.635918	127.455140
min	1.000000	58.000000	61.000000
25%	1.000000	361.000000	403.750000
50%	1.000000	432.500000	463.000000
75%	1.000000	490.000000	526.000000
max	3.000000	796.000000	961.000000

	weight_pounds	bmi	fat
count	67.000000	67.000000	2.00000
mean	158.811801	25.185224	23.50000
std	30.695415	3.066963	2.12132
min	115.963147	21.450001	22.00000
25%	135.363832	23.959999	22.75000
50%	137.788914	24.389999	23.50000
75%	187.503152	25.559999	24.25000
max	294.317120	47.540001	25.00000

#percentages
print("percent with 0 logged activities:","{:.2%}".format(np.sum(updated_daily_activity["logged_activities_distance"] == 0) / updated_daily_activity["logged_activities_distance"].count()))
print("percent with 0 distance:","{:.2%}".format(np.sum(updated_daily_activity["total_distance"] == 0) / updated_daily_activity["total_distance"].count()))
print("percent users particpated in weight logging:","{:.2%}".format(updated_weight_log["id"].nunique() / updated_daily_activity["id"].nunique()))
print("percent users particpated in sleep logging:","{:.2%}".format(updated_daily_sleep["id"].nunique() / updated_daily_activity["id"].nunique()))

percent with 0 logged activities: 96.60%
percent with 0 distance: 8.30%
percent users particpated in weight logging: 24.24%
percent users particpated in sleep logging: 72.73%

#merge daily activity + daily sleep outer join
combined_sleep_activity = pd.merge(updated_daily_sleep, updated_daily_activity, how="right", on=["id","activity_date"])

#merge all daily data outer join
combined_all_daily = pd.merge(combined_sleep_activity, updated_weight_log, how="left", on=["id"])

combined_sleep_activity.head()
combined_all_daily.head()

	id	activity_date	total_sleep_records	total_minutes_asleep	total_time_in_bed	weekday_x	total_steps	total_distance	tracker_distance	logged_activities_distance	very_active_distance	moderately_active_distance	light_active_distance	sedentary_active_distance	very_active_minutes	fairly_active_minutes	lightly_active_minutes	sedentary_minutes	calories	weekday_y
0	1503960366	2016-04-12	1.0	327.0	346.0	Tuesday	13162	8.50	8.50	0.0	1.88	0.55	6.06	0.0	25	13	328	728	1985	Tuesday
1	1927972279	2016-04-12	3.0	750.0	775.0	Tuesday	678	0.47	0.47	0.0	0.00	0.00	0.47	0.0	0	0	55	734	2220	Tuesday
2	2022484408	2016-04-12	NaN	NaN	NaN	NaN	11875	8.34	8.34	0.0	3.31	0.77	4.26	0.0	42	14	227	1157	2390	Tuesday
3	8877689391	2016-04-12	NaN	NaN	NaN	NaN	23186	20.40	20.40	0.0	12.22	0.34	7.82	0.0	85	7	312	1036	3921	Tuesday
4	2026352035	2016-04-12	1.0	503.0	546.0	Tuesday	4414	2.74	2.74	0.0	0.19	0.35	2.20	0.0	3	8	181	706	1459	Tuesday

	id	activity_date_x	total_sleep_records	total_minutes_asleep	total_time_in_bed	weekday_x	total_steps	total_distance	tracker_distance	logged_activities_distance	...	weekday_y	activity_date_y	weight_kg	weight_pounds	fat	bmi	is_manual_report	log_id	weekday	report_type
0	1503960366	2016-04-12	1.0	327.0	346.0	Tuesday	13162	8.50	8.50	0.0	...	Tuesday	2016-05-02 23:59:59	52.599998	115.963147	22.0	22.650000	True	1.462234e+12	Monday	Manual
1	1503960366	2016-04-12	1.0	327.0	346.0	Tuesday	13162	8.50	8.50	0.0	...	Tuesday	2016-05-03 23:59:59	52.599998	115.963147	NaN	22.650000	True	1.462320e+12	Tuesday	Manual
2	1927972279	2016-04-12	3.0	750.0	775.0	Tuesday	678	0.47	0.47	0.0	...	Tuesday	2016-04-13 01:08:52	133.500000	294.317120	NaN	47.540001	False	1.460510e+12	Wednesday	Automated
3	2022484408	2016-04-12	NaN	NaN	NaN	NaN	11875	8.34	8.34	0.0	...	Tuesday	NaT	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4	8877689391	2016-04-12	NaN	NaN	NaN	NaN	23186	20.40	20.40	0.0	...	Tuesday	2016-04-12 06:47:11	85.800003	189.156628	NaN	25.680000	False	1.460444e+12	Tuesday	Automated

5 rows × 29 columns

combined_sleep_activity["id"].nunique()
combined_all_daily["id"].nunique()

33

33

average_steps_overall = updated_daily_activity["total_steps"].mean()

---

Step 5 - Share

Visualizations

Graphs Categorized by Weekday

avg_steps_day = updated_daily_activity.groupby(["weekday"])["total_steps"].mean().sort_values(ascending = False)
avg_steps_day

weekday
Saturday     8152.975806
Tuesday      8125.006579
Monday       7780.866667
Wednesday    7559.373333
Friday       7448.230159
Thursday     7405.836735
Sunday       6933.231405
Name: total_steps, dtype: float64

avg_steps_day.plot(kind="bar")
plt.title("Average Steps per Day")
plt.xlabel("Day")
plt.ylabel("Averge Steps")
plt.axhline(10000, color='green')

<AxesSubplot:xlabel='weekday'>

Text(0.5, 1.0, 'Average Steps per Day')

Text(0.5, 0, 'Day')

Text(0, 0.5, 'Averge Steps')

<matplotlib.lines.Line2D at 0x20c0cff00d0>

#Average Sleep per Day
avg_sleep_day = updated_daily_sleep.groupby(["weekday"])["total_minutes_asleep"].mean().sort_values(ascending = False)
avg_sleep_day

weekday
Sunday       452.745455
Wednesday    434.681818
Monday       419.500000
Saturday     419.070175
Friday       405.421053
Tuesday      404.538462
Thursday     401.296875
Name: total_minutes_asleep, dtype: float64

avg_sleep_day.plot(kind="bar")
plt.title("Average Sleep per Day")
plt.xlabel("Day")
plt.ylabel("Average Minutes")
plt.axhline(480, color='green')

<AxesSubplot:xlabel='weekday'>

Text(0.5, 1.0, 'Average Sleep per Day')

Text(0.5, 0, 'Day')

Text(0, 0.5, 'Average Minutes')

<matplotlib.lines.Line2D at 0x20c0fd411f0>

#Daily Weight Log
updated_weight_log['weekday'].value_counts().plot(kind='bar')
plt.title("Daily Weight Logs")
plt.xlabel("Day")
plt.ylabel("Total Log Count")

<AxesSubplot:>

Text(0.5, 1.0, 'Daily Weight Logs')

Text(0.5, 0, 'Day')

Text(0, 0.5, 'Total Log Count')

#average minutes per level
dataplot = updated_daily_activity[["weekday","sedentary_minutes","lightly_active_minutes", "fairly_active_minutes","very_active_minutes"]]
dataplot = dataplot.groupby('weekday').agg('mean')
dataplot.plot(kind="bar")
plt.title("Activity Intensity per Day")
plt.xlabel("Day")
plt.ylabel("Average Minutes")

<AxesSubplot:xlabel='weekday'>

Text(0.5, 1.0, 'Activity Intensity per Day')

Text(0.5, 0, 'Day')

Text(0, 0.5, 'Average Minutes')

Relationships and Trend Lines

#relationship between total_time_in_bed and total_minutes_asleep
sns.regplot(x="total_time_in_bed", y="total_minutes_asleep", data=updated_daily_sleep, scatter_kws={"color":"black"}, line_kws={"color":"blue"})

plt.title("Time in Bed vs Time Asleep") 
plt.ylabel("Asleep")
plt.xlabel("In Bed") 
plt.axhline(480, color='green')
plt.text(600,800,'8 hour recommended', color='green')
plt.show()

<AxesSubplot:xlabel='total_time_in_bed', ylabel='total_minutes_asleep'>

Text(0.5, 1.0, 'Time in Bed vs Time Asleep')

Text(0, 0.5, 'Asleep')

Text(0.5, 0, 'In Bed')

<matplotlib.lines.Line2D at 0x20c066cb0d0>

Text(600, 800, '8 hour recommended')

#relationship between total_steps and sedentary_minutes
sns.regplot(x="total_steps", y="sedentary_minutes", data=updated_daily_activity, scatter_kws={"color":"black"}, line_kws={"color":"blue"})
plt.title("Steps vs Sedentary Minutes") 
plt.xlabel("Total Steps") 
plt.ylabel("Sedentary Minutes")
plt.show()

<AxesSubplot:xlabel='total_steps', ylabel='sedentary_minutes'>

Text(0.5, 1.0, 'Steps vs Sedentary Minutes')

Text(0.5, 0, 'Total Steps')

Text(0, 0.5, 'Sedentary Minutes')

#total steps vs total minutes asleep
sns.regplot(x="total_minutes_asleep", y="total_steps", data=combined_sleep_activity, scatter_kws={"color":"black"}, line_kws={"color":"blue"})

plt.title("Steps vs Sleep") 
plt.xlabel("Minutes Asleep") 
plt.ylabel("Total Steps")
plt.axvline(480, color='green')
plt.text(600,20000,'8 hour recommended', color='green')
plt.show()

<AxesSubplot:xlabel='total_minutes_asleep', ylabel='total_steps'>

Text(0.5, 1.0, 'Steps vs Sleep')

Text(0.5, 0, 'Minutes Asleep')

Text(0, 0.5, 'Total Steps')

<matplotlib.lines.Line2D at 0x20c066ccd90>

Text(600, 20000, '8 hour recommended')

#total minutes asleep vs calories
sns.regplot(x="total_minutes_asleep", y="calories", data=combined_sleep_activity, scatter_kws={"color":"black"}, line_kws={"color":"blue"})

plt.title("Sleep vs Calories") 
plt.xlabel("Minutes Asleep") 
plt.ylabel("Calories")
plt.axvline(480, color='green')
plt.text(600,4000,'8 hour recommended', color='green')
plt.show()

<AxesSubplot:xlabel='total_minutes_asleep', ylabel='calories'>

Text(0.5, 1.0, 'Sleep vs Calories')

Text(0.5, 0, 'Minutes Asleep')

Text(0, 0.5, 'Calories')

<matplotlib.lines.Line2D at 0x20c066cb220>

Text(600, 4000, '8 hour recommended')

Participation Graphs

#step count per user

user_total_steps = updated_daily_activity.groupby('id')[["total_steps"]].sum().sort_values('total_steps', ascending = False)

top_10_users = user_total_steps[:10]

user_avg_steps = updated_daily_activity.groupby('id')[["total_steps"]].mean().sort_values('total_steps', ascending = False)

top_10_users_mean = user_avg_steps[:10]

user_avg_steps.plot(kind="bar")
plt.title("Average Steps per User")
plt.xlabel("User")
plt.ylabel("Average Steps")
plt.axhline(10000, color='green')

<AxesSubplot:xlabel='id'>

Text(0.5, 1.0, 'Average Steps per User')

Text(0.5, 0, 'User')

Text(0, 0.5, 'Average Steps')

<matplotlib.lines.Line2D at 0x20c05cbc700>

#Participation
updated_daily_activity['id'].value_counts().plot(kind='bar')
plt.title("Physical Activity Participant Frequency")
plt.xlabel("User ID")
plt.ylabel("Total Log Count")

<AxesSubplot:>

Text(0.5, 1.0, 'Physical Activity Participant Frequency')

Text(0.5, 0, 'User ID')

Text(0, 0.5, 'Total Log Count')

updated_daily_sleep['id'].value_counts().plot(kind='bar')
plt.title("Sleep Participant Frequency")
plt.xlabel("User ID")
plt.ylabel("Total Log Count")

<AxesSubplot:>

Text(0.5, 1.0, 'Sleep Participant Frequency')

Text(0.5, 0, 'User ID')

Text(0, 0.5, 'Total Log Count')

updated_weight_log['id'].value_counts().plot(kind='bar')
plt.title("Weight Log Particpant Frequency")
plt.xlabel("User ID")
plt.ylabel("Total Log Count")

<AxesSubplot:>

Text(0.5, 1.0, 'Weight Log Particpant Frequency')

Text(0.5, 0, 'User ID')

Text(0, 0.5, 'Total Log Count')

#all particpation activity
record_type = ["Physical", "Sleep", "Weight"]
distinct_user_count = [updated_daily_activity["id"].nunique(),updated_daily_sleep["id"].nunique(),updated_weight_log["id"].nunique()]

fig1, ax1 = plt.subplots()
ax1.pie(distinct_user_count, labels=record_type, autopct='%1.1f%%', startangle=90)
ax1.axis('equal')  # Equal aspect ratio ensures that pie is drawn as a circle.

ax1.set_title("All Participation Activity")

plt.show()

([<matplotlib.patches.Wedge at 0x20c7a98f310>,
  <matplotlib.patches.Wedge at 0x20c7a98ff10>,
  <matplotlib.patches.Wedge at 0x20c7a52dee0>],
 [Text(-1.0996788130378836, -0.026580221135116266, 'Physical'),
  Text(1.0285179484228668, -0.3900651609308596, 'Sleep'),
  Text(0.41480408775850713, 1.0187922107961136, 'Weight')],
 [Text(-0.5998248071115728, -0.014498302437336144, '50.8%'),
  Text(0.5610097900488363, -0.21276281505319614, '36.9%'),
  Text(0.22625677514100387, 0.5557048422524256, '12.3%')])

(-1.1039992533980778, 1.11951741980678, -1.105538688392744, 1.10026377876888)

Text(0.5, 1.0, 'All Participation Activity')

Manual vs Automated Preferences

#percent of total records of 0 logged activities
manual_physical = np.sum(updated_daily_activity["logged_activities_distance"] != 0)
automated_physical = np.sum(updated_daily_activity["logged_activities_distance"] == 0)
record_type_count = [manual_physical,automated_physical]

labels = 'Manual', 'Automated'
fig1, ax1 = plt.subplots()
ax1.pie(record_type_count, labels=labels, autopct='%1.1f%%')
ax1.axis('equal')  # Equal aspect ratio ensures that pie is drawn as a circle.

ax1.set_title("Manual vs Automated Physical Activity Logs")

plt.show()

([<matplotlib.patches.Wedge at 0x20c7a0ec760>,
  <matplotlib.patches.Wedge at 0x20c7a08f520>],
 [Text(1.0937151825418197, 0.11741848013628155, 'Manual'),
  Text(-1.0937151770450626, -0.11741853133678419, 'Automated')],
 [Text(0.5965719177500833, 0.06404644371069902, '3.4%'),
  Text(-0.5965719147518523, -0.06404647163824591, '96.6%')])

(-1.1082151582541035,
 1.1003912104513915,
 -1.1059873583562982,
 1.1080560958448729)

Text(0.5, 1.0, 'Manual vs Automated Physical Activity Logs')

Time Lapse

#Shows participants log days
hue_order = ['Manual', 'Automated']
sns.scatterplot(x="activity_date", y="id", data=updated_weight_log, hue = "report_type", hue_order=hue_order )

plt.title("Weight Log Consistency") 
plt.xlabel("Date of Log") 
plt.ylabel("User ID")

<AxesSubplot:xlabel='activity_date', ylabel='id'>

Text(0.5, 1.0, 'Weight Log Consistency')

Text(0.5, 0, 'Date of Log')

Text(0, 0.5, 'User ID')

g = sns.FacetGrid(updated_weight_log, col ="id", col_wrap=4, col_order = updated_weight_log['id'].value_counts().index)
g.map(sns.lineplot, "activity_date", "weight_pounds")
g.set_xticklabels(rotation=45)

plt.title("Weight Change") 
plt.xlabel("Date of Log") 
plt.ylabel("Weight in Pounds")

<seaborn.axisgrid.FacetGrid at 0x20c7965ea60>

<seaborn.axisgrid.FacetGrid at 0x20c7965ea60>

Text(0.5, 1.0, 'Weight Change')

Text(0.5, 6.800000000000011, 'Date of Log')

Text(669.575, 0.5, 'Weight in Pounds')

#Steps over Time
sns.lineplot(data = updated_daily_activity, x="activity_date", y="total_steps")

plt.title("User Activitiy")
plt.xlabel("Date")
plt.ylabel("Steps Taken")
plt.xticks(rotation=45)
plt.show()

<AxesSubplot:xlabel='activity_date', ylabel='total_steps'>

Text(0.5, 1.0, 'User Activitiy')

Text(0.5, 0, 'Date')

Text(0, 0.5, 'Steps Taken')

(array([16904., 16908., 16912., 16916., 16920., 16922., 16926., 16930.,
        16934.]),
 [Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, '')])

#Steps over Time
sns.lineplot(data = updated_daily_activity, x="activity_date", y="total_steps", hue="id")

plt.title("User Activitiy")
plt.xlabel("Date")
plt.ylabel("Steps Taken")
plt.xticks(rotation=45)
plt.show()

<AxesSubplot:xlabel='activity_date', ylabel='total_steps'>

Text(0.5, 1.0, 'User Activitiy')

Text(0.5, 0, 'Date')

Text(0, 0.5, 'Steps Taken')

(array([16904., 16908., 16912., 16916., 16920., 16922., 16926., 16930.,
        16934.]),
 [Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, ''),
  Text(0, 0, '')])

Observations from Analysis

Participation

• Users were inconsistent in tracking across all activities
• Of 33 participants total
  • 33 participated in physical activity logging
    • Majority logged everyday, however not all did.
  • 24 participated in sleep activity logging
    • Majority did not log everyday
  • 8 participated in weight logging activity
    • Majority did not log everyday

Weekday Analysis

• Saturday has the highest average step count. Sunday has the lowest. Not too much deviation as the week progresses
• Sunday has the highest average sleep count. Thursday has the lowest. Not too much deviation as the week progresses
• Monday and Wednesday are the most active days for Weight Log recording. Friday and Saturday are the least active days. Slight trend of lesser activity as the week progresses

Daily Sleep Activity

• The average sleep time is under the recommended 8 hours every day
• There is a trend that users spend roughly the same amount of time in bed as they spend actually sleeping. Meaning they fall asleep and stay asleep, and do not lie in bed awake for long. There are some outliers here that spend a lot of time awake in bed.

Daily Physical Activity

• The average steps taken is under the recommended 10,000 every day
• As total steps increases, the amount of sedentary minutes decreases, as one may assume.
• Users are are mostly sedentary during each day
• Users are consistently "lightly active" over moderately and very active.

Coorelation between Physical and Sleep Activity

• There is a trend showing the more sleep you get, the less steps you take.
  • This could mean that there is less time to be active if you sleep longer
• There is not a strong correlation between the amount of sleep and the amount of calories burned.

Weight Log Analysis

• 2 of 8 participants accounted for most of the logs, and the same 2 were most consistent in logging nearly everyday.
• The remaining 6 participants were not consistent and randomly logged.
• There were no drastic changes in weight. The study being only 30 days may account for this.

Manual vs Automated Logging

• 4 out of 33 manually logged physical activity.
• It is inferred that automated logging is the preference over manual logging with physical activity.

---

Step 6 - Act

Final Conclusions

• Overall, users were not consistent with tracking physical activity, sleep activity, and logging weight. Potential reasons:
  • Users were forgetting to wear the device during physical activity and during sleep
  • Users were not wanting to wear the device due to comfortability
  • Users were forgetting to manually log activity
  • Users were not finding the tracking process easy and convenient
• Users were more consistent with tracking physical activities, which could indicate that the device is not comfortable enough to wear during sleep, and that the device does not allow easy weight logging entries.
• 96% of logged physical activity records are automated. Meaning users rely on activity being automatically detected and logged, and not manually entered.

Recommendations

To meet the goal to empower women with knowledge about their own health and habits, the following recommendations for marketing are to:

Create and Advertise

• A more comfortable device that allows all-day wear for consistent tracking which leads to more complete health data
• Daily challenges to provide incentive to stay consistent with all activities - physical, sleep, and logging each day of the week
  • Challenge to take 10,000 steps
  • Challenge to sleep 8 hours
• An easy and convenient tracking process focused on automated activity tracking, and less need for manual intervention.
• Reminders to log activity

Next Steps

• More forward with recommendations with the insights gained from the limited data
  or

• Expand upon the limited data and timeframe by conducting another survey to gather specific and categorical data that targets the female audience. And to ask questions around specific goals that they have with their health and fitness tracking.