The smart phone activity distribution shows that the data is almost equally distributed and balanced, but laying takes the lead!
Let's investigate the data more...
if the value is < -0.8 - standing, sitting or laying
if the value is > -0.6 - walking, walking down or walking upstairs
if the value is > 0.0 - walking downstairs
Can we differentiate dynamic from static activities?
To better comprehend the static and dynamic activities of humans, we are utilizing the "tBodyAccMagmean" (tBody acceleration magnitude feature mean value) function to produce the graph.
Activity is laying if angleX, gravityMean is greater than 0.
With just one if else statement, we can categorize every data piece associated with the Laying activity.
Have you ever wondered what a packed-bubble plot looks like?
Hint : A dart board
This "dart board" represents how the data points which is the distribution of the different activities are stacked in a 2D plot.
Scroll down to see how the same data looks like in 3D.
This is a 3-D plot of the points that represent the six different activities suspended in the 3 axes - X,Y & Z
You can zoom in, zoom out, hover over a point to understand it better or interact with it however you'd like to.
Go ahead and try it!
We have used a visualization tool in Python which uses a statistical method called "t-sne", that basically separates data points for classification, and helps us understand complex, high-dimensional data by reducing and projecting it in a low dimensional space such as in 2D or 3D.
The carousel on the left, shows, slide by slide, how t-sne classifies the different activities clearly which makes it easy to analyze the data further.
Keep swiping right to see how the perplexity affects the classification