Interactive Data Exploration, DIKU 2015

Interesting Data Points

Highlight singled out index finger (point 39) or singled out ring finger (point 37).

Highlight spread fingers (point 30) or closed fingers (point 35).

2.1 Principle Components

Using the Principle Component Analysis (PCA), the two most significant principle components (PC1 and PC2) are a combination of the most distinguishing elements describing the outline of a hand. This means that only a few principle components, in this case two, can be used to separate the data (i.e. 2D coordinates for the outline drawing of a hand) in a meaningful way quite well.

In order to illustrate this, we can see two selectable principle components on two axes of a two-dimensional scatterplot. These data points correspond to a whole dataset of coordinates that outline the drawing of a hand.

2.2 Spread of Fingers

It can be seen that the most significant principle component PC1 gives an accurate indication for the spread of the fingers. The higher the first principle component is, the less spread out the fingers are. For instance, the data point 30, with the lowest PC1 value among all data points of -0.49, corresponds to a hand with all fingers far spread. In contrast the data point 35 has the highest PC1 value of 0.62 and corresponds to a hand with all fingers closed.

If the second most significant principle component (PC2) is taken into account as well, we can see a different pattern emerge. The data point 39 (PC1: -0.42, PC2: -0.43) shows the index finger singled out with all other fingers spread away from it. Again, in contrast, we can have a look at data point 37 (PC1: 0.12, PC2: 0.47) which shows the ring finger singled out and all the other fingers spread away from it.

3.1 Bogdan

• Add coloring based on clustering. I have been using this implementation of the k-means algorithm.
• Zooming in and out on scatter plot

3.2 Hauke

• Display the datafile row index as a text-label or a tooltip when the mouse hovers over its point in the PCA panel.
• Support changing which two PCA-variables are displayed in panel two (with a nice transition).
• Enable users to share links that highlight a point in the defined dimensions.

3.3 Marco

• Connect a piece of text in the discussion with the visualization. For example, when the mouse hovers over the discussion about an outlier, then the outlier gets highlighed in the visualization. In this case one has to click on the text part, as can be seen in the "Interesting Data Points" section.