4.4.3 Clean and rename columns

To ensure consistency in column names across dataset and facilitate their later on,

we will convert all column names to lowercase, thereby ensuring that they adhere to the

correct syntax and format

4.4.4 Consistency Of Date and Time Columns

After confirming and converting our column names to lowercase, the next step is to clean the date-time format for the Daily_Activity and Daily_Sleep data frames. This is important as we plan to merge the two data frames. As the time is not relevant in the Daily_Sleep data frame, we will use the as_date function instead of as_datetime.

We will check our cleaned datasets 

head(Daily_Activity)

head(Daily_Sleep

We will convert the date string to date-time for our Hourly_Steps data set.

4.5 Merging Datasets 

We will merge Daily_Activity and Daily_Sleep to see any correlation between variables by using ID and date as their primary keys.

Now that we have added a new column indicating the user_ type, we can create a data frame that shows the percentage of users in each category, which will allow us to visualize the distribution more clearly on a graph.

Now we can see that users are fairly distributed by their activity considering the daily amount of steps. We can determine that based on users activity all kinds of users wear smart-devices.

5.2 Steps and Minutes Asleep Per Weekday 

We are interested in determining the most active and restful days of the week for our users, as well as whether they meet recommended activity and sleep goals. To this end, we are using our date column to determine the day of the week for each record, and calculating the average number of steps and minutes of sleep for each weekday.

Based on the graphs, we can observe that the users generally meet the recommended amount of steps per day of 7500, except for Sundays. However, the users do not meet the recommended amount of sleep, which is 8 hours per night.

5.3 Hourly Steps Throughout The Day 

To further explore our analysis, we aim to determine the specific times of day when users are most active. This can be achieved by utilizing the Hourly_Steps data frame and parsing out the date_time column.

Based on the data analysis, it appears that users tend to be more active between the hours of 8am and 7pm. Specifically, they seem to take more steps during lunchtime, from 12pm to 2pm, as well as in the evenings, between 5pm and 7pm.

5.4 Correlations 

Certainly, we can examine whether there is a correlation between different variables. 

• Daily steps and daily sleep

• Daily steps and calories

Based on the analysis, it appears that there is no correlation between daily activity level based on steps and the amount of minutes users sleep a day. This indicates that one variable does not affect the other, and they are independent of each other.

On the other hand, we observe a positive correlation between daily steps and calories burned. This suggests that as the number of steps walked increases, the number of calories burned also tends to increase. This relationship is likely due to the fact that physical activity leads to an increase in energy expenditure, which in turn can lead to a higher calorie burn. However, it is important to note that correlation does not necessarily imply causation, and further analysis may be needed to establish a cause-and-effect relationship between these variables.

5.5 Use of smart device 

         5.5.1 Days Used Smart Device

To determine the frequency of device usage among the users in our sample, we can create a new data frame that groups the data by Id and calculates the number of days each user used their device. Then, we can use this information to classify each user into one of three categories based on their level of usage:

high use: users who used their device between 21 and 31 days

moderate use: users who used their device between 10 and 20 days

low use: users who used their device between 1 and 10 days

Here's an example of how we can create the new data frame and classification column with the classification explained above.

It seems that the majority of users in our sample use their smart devices frequently, with 50% of users using their device between 21 to 31 days in a 31-day interval. This is a good indication that there is a strong demand for smart devices and their features among our target audience.

However, it's also important to note that a significant portion of our sample (38%) uses their device very rarely, which may suggest that there is some room for improvement in terms of encouraging device usage and increasing user engagement. This information could be useful in informing our marketing and product development strategies, as we may want to focus on features and promotions that are specifically designed to increase user engagement and device usage. The moderate usage group of 12% also provides valuable insight and could be an important target audience for certain marketing efforts.

5.5.2 Time Used Smart Device 

Merging the Daily_Use data frame and Daily_Activity data frame will allow us to see how many minutes users wear their device per day and filter the results by daily use of the device.

We need to generate a new data frame that computes the total number of minutes each user wore the device per day, and classify the data into one of three categories:

1. All day: if the device was worn for the entire day.

2. More than half day: if the device was worn for more than half of the day.

3. Less than half day: if the device was worn for less than half of the day.

To facilitate the visualization of our results, we will create four new data frames. These data frames will allow us to compare the total usage with the usage by category and to make meaningful conclusions.

The first data frame will display the total number of users, and the percentage of time the device was worn, classified by the three categories we created (All day, More than half day, Less than half day). By calculating the total minutes worn by each user per day, we will determine the percentage of minutes worn in each of the three categories. This data frame will have columns for the total number of users, the percentage of users wearing the device all day, the percentage of users wearing the device more than half day, and the percentage of users wearing the device less than half day.

The remaining three data frames will be filtered by category of daily users, to better understand the differences in usage patterns. The second data frame will show data only for users who wore the device All day. The third data frame will show data only for users who wore the device More than half day. The fourth data frame will show data only for users who wore the device Less than half day. These data frames will have columns for the total number of users and the total number of minutes worn by each user in each category.

By creating these four data frames, we can compare the overall usage with the usage by category, and draw meaningful conclusions about device usage patterns. We can arrange these data frames on the same visualization to facilitate comparisons and analysis.

head(minutes_worn_percent)

head(minutes_worn_highuse)

head(minutes_worn_moduse)

head(minutes_worn_lowuse)

Now that we have created the four data frames and organized the worn level categories, we can visualize our results using the following plots. To facilitate comparisons and analysis, all plots have been arranged together.

Our plots show that 36% of all users wear the device all day, while 60% wear it for more than half the day, and only 4% wear it for less than half the day. When we filter the total users by the number of days they used the device and the amount of time they wore it each day, we can observe some interesting patterns.

For users who used the device between 21 and 31 days (high use), only 6.8% wear it all day. The vast majority (88.9%) wear it for more than half the day but not all day.

Moderate users, who used the device between 10 and 20 days, tend to wear it less on a daily basis.

Interestingly, low users who used the device between 1 and 10 days tend to wear it for longer periods on the days they do use it.

These observations highlight the different usage patterns of users across different time frames and usage levels. By analyzing these patterns, we can gain insights into how users interact with the device and use this information to inform our future decisions and strategies.

6. Conclusion (Act Phase) 

Bellabeat's mission is to empower women by providing them with the data to discover themselves. Based on the results of our analysis, I would recommend that we utilize our own tracking data for further analysis to better understand our target audience. The datasets we used had a small sample and lacked demographic details of users, which may have introduced biases in our analysis. As our main target audience is young and adult women, it is important that we continue to identify trends and insights that can inform a marketing strategy specifically tailored to them.

By utilizing our own tracking data, we can gather more accurate and representative information about our users' behaviors and preferences. This will allow us to develop more personalized and effective marketing campaigns that resonate with our target audience and help them discover the full potential of our products. Overall, our goal is to continue advancing Bellabeat's mission of empowering women through data-driven insights and innovation.