Qiao | Byte3 Mobile data

This is a topic about mobile data

Main Topic: what factors will have a significant effect on battery usage and how do they affect battery usage?

1. Definition: measurement of the battery usage

Here I use battery charge/discharge efficiency as a metric to measure battery usage

Here I use battery charge/discharge efficiency as a metric to measure battery usage

Battery efficiency = Δtime/ΔBattery

For example, battery discharges from 90% to 24%, and it takes 3hours for it. In this case, Battery efficiency = 3*60*60*1000 / (90-24)

2. Collecting Data-Battery consumption

Here I use data from tables "Battery", "Battery_discharge", "Battery_charge" as a starting point, which is collected thorugh Aware.

After roughly cleaning the data, and dropped some irrelevant column, the datasets are visuallized as follows:

Figure 2.1 Battery Consumption

Figure 2.1.2 Battery charge pattern

Figure 2.2 Battery discharge

Figure 2.2 Battery discharge

Conclusion: the battery is consumed mostly the same way, in which the battery efficiency would have to be around the same patter.

In addition, there are obivous flaws shown on the graph that needs to be cleaned.

3. Collecting data-Battery efficieny

Here I use data from tables to calculate the battery efficiency according to the equation at the begining.

After calculation battery charge/discharge efficiency is shown as follows:

figure3.1 Battery charge efficiency:

figure3.2 Battery discharge efficiency

Conclusion: In most part, the battery efficiency is fluctuating with a certain range. However, there are some obvious outliers that needs furthur explore.

4. Cleaning the data

Before I start cleaning the data, I took a closer look at the outliers both in charge efficiency and discharge efficiency.

After get a better understaning on the documentation that Aware provide on its website, I finally found the reason behind outliers.

There are negtive number in efficency tables, meaning when you charge your phone you are losing battery sharply, which is definitely impossible. And it turns out it is the result of abnormal behavior happened to the Aware. Sometimes I will quit Aware in order to save battery, thus it happend to Aware that it just record 2 unassociated number and place them in the same row.

And for those 0 value number, it is due to low battery mode or Aware error.

After cleaning the irrelevant data, the figures are as follows

Figure 4.2 Battery discharge efficiency(After cleaning)

Figure 4.2 Battery discharge efficiency(After cleaning)

5. Exploring the data

For the battery discharge efficiency, the average is 739319.4701 milliseconds/per%

For the battery charge efficiency, the average is 73840.84029 milliseconds/per%

5.1 Explore the relationships between locations and battery efficiency.

Figure 5.1.1 locations



Figure 5.2.1 battery charge efficiency distribution geomatically

After cleaning the data again to match the time when battery is charged, the result is as follows. I use average battery charge efficiency to differentiate the location points. Location points that have higher battery efficiency than average are green, and points that have lower battery efficiency is as above:

(Due to asynchronous timestamps issue, there is a lag on the map--even the battery is not being charged, the points are still displayed on the map. But for those places which is away from my home and school, the points represents the battery efficiency at that time.)

Conclusion: for those points around my house(at the cross of Morewood Ave and Center Ave) and school, the points represents the real battery charge efficiency. From the map, we can find that it is more likely that charge efficiency is higher in my house than in my school. The reason behind that is because I don't use my phone when charging it when I am at home. For example, I will use my laptop instead of phone when I am home and just leave my phone charged aside. But When I am in school, I will charge my phone mostly because I found it is dying when I have to use it, meaning I will have to consume battery when I charge it.

Figure 5.3.1 battery discharge efficiency distribution geomatically

After cleaning the data again to match the time when battery discharges, the result is as follows. I use average battery discharge efficiency to differentiate the location points. Location points that have higher battery discharge efficiency than average are green, and points that have lower battery efficiency is above:

Here I found three interesting insights:
1. Around Allegheny Center on the map, all points are yellow. It turns out it is because I used my phone to navigate that time.
2. Those places where I got low battery discharge efficiency is usually where I have rely on my phone to navigate or to spend spend spare time.
3. On the commuting route, there are high and low battery discharge efficieny existing at the same time. It is because that I listen to music on the bus usually but sometimes got nothing to do on the bus as well.

5.2 Explore the relationships between screen status and battery consumption.

Figure 5.2.1 screen staus (screen status: 0=off, 1=on, 2=locked, 3=unlocked)



As Figuire 5.2.1 shows, mostly I will change the screen status from off to unlocked directly without going through loked, since the 'Touch ID' enable me to unlock my phone from the status when the screen is off. Sometimes screen will jump from off to locked as well, this is due to that IOS now enable users to use a lift gesture to wake up the phone screen, which take place very frequently.

Figure 5.2.2 time duration of each screen status

Figure 5.2.2 suggests my typical phone usage pattern. I will keep my phone on mostly won't exceed around 40minutes(2430sec), but there is an outliers in on status paired with off shown in the chart, which reveal an unnormal usage. The reason behind that is I forget to bring my laptop and have to rely on my phone to work at that time.

Figure 5.2.3 Relationship between battery discharge and screen status

Figure 5.2.3 indicates a close association, in which most screen status switches correspond to battery status swithches. This pattern reveals actually a very apparent phenomenon--I usually charge and discharge my phone when it is on. When I turn on screen and found the battery is low, I will immediately get it charged--thus these two are closely associated.

Figure 5.2.4 Relationship between locations and screen status((screen status: 0=off, 1=on, 2=locked, 3=unlocked))

Figure 5.2.4 indicates the relationships between locations and screen status. Yellow points represent places where screen is off, the greens are "on", purples are locked, reds are unlocked. Mostly they are distributed randomly due to the specific events at that time. But when it is some places where I will stay static, for example, bus stops, campus, and house, the red points are increasingly intensive.

Average Daily Consumption, Per Person

Speed
Pause
Slow
Medium
Fast
Year
1970