这这一课程中，我们将学习数据挖掘的基本概念及其基础的方法和应用，然后深入到数据挖掘的子领域——模式发现中，深入学习模式发现的概念、方法，及应用。我们也将介绍基于模式进行分类的方法以及一些模式发现有趣的应用。这一课程将给你提供学习技能和实践的机会，将可扩展的模式发现方法应用在在大体量交易数据上，讨论模式评估指标，以及学习用于挖掘各类不同的模式、序列模式，以及子图模式的方法。

Loading...

来自 University of Illinois at Urbana-Champaign 的课程

数据可视化

595 个评分

这这一课程中，我们将学习数据挖掘的基本概念及其基础的方法和应用，然后深入到数据挖掘的子领域——模式发现中，深入学习模式发现的概念、方法，及应用。我们也将介绍基于模式进行分类的方法以及一些模式发现有趣的应用。这一课程将给你提供学习技能和实践的机会，将可扩展的模式发现方法应用在在大体量交易数据上，讨论模式评估指标，以及学习用于挖掘各类不同的模式、序列模式，以及子图模式的方法。

从本节课中

Week 2: Visualization of Numerical Data

In this week's module, you will start to think about how to visualize data effectively. This will include assigning data to appropriate chart elements, using glyphs, parallel coordinates, and streamgraphs, as well as implementing principles of design and color to make your visualizations more engaging and effective.

- John C. HartProfessor of Computer Science

Department of Computer Science

[SOUND] So

often we want to visualize multiple variables simultaneously.

And so we generate stacked graphs that can display several variables

changing along the same dimension.

For a bar chart, we may want to look at more than just a single quantitative

dependent variable measured across an independent variable dimension.

One way of doing this is with a stacked bar chart.

In this case we can have two or more dependent variables plotted.

In this case, the blue bars are representing one dependent variable and

the red bars are representing a second dependent variable.

And so we're using q to represent different nominal value, basically

the second dependent variable is different than the first dependent variable.

And then the total amount of the contribution of both variables

is represented by the height of the top of these stacked bars.

And the relative proportion is being represented by

the proportion of the bar being one color versus another color.

And if we want to emphasize that relative proportion.

We can have the bars stacked to 100% and

just plot the percentage of the contribution of one variable versus

the other variable as the difference in the colors of the bars.

This kind of relative contribution is also represented in a pie chart but

in a pie chart you're representing the contribution based on angle and

a little bit by area than on position and length.

Here in a pie chart the relative proportion of the blue category

versus the orange category versus the green category versus the yellow category

is being represented by the angle as the angle's sum to 360 degrees.

It's also proportional to the area represented by each of these four regions.

That indication can be degraded somewhat if you try to

show a pie chart in three dimensions.

And that's because in three dimensions we are still perceiving

a three dimensional scene as a two dimensional image.

And the visual cues we're using include foreshortening and perspective.

And those cues can get in the way of our ability to perceive a region, say in

the foreground, as being larger or smaller than a region in the background because of

our expectation that perspective is going to make regions in the foreground larger.

Or at least appear larger and regions in the background appear smaller,

even if it's just a orthographic non-perspective projection.

So, it's best not to use three dimensions for pie charts and, in fact,

it's better just to use a stacked bar chart instead of a pie chart.

In this case, each one of these bars can represent a separate pie chart

because it's representing the separate relative contribution to the whole of

multiple dependent variables, as they vary across some independent variable.

Also, stacking order matters.

I've got the same data plotted in the left and in the right.

In the left, I've plotted the blue variable first.

And then I've stacked the red variable on top of that.

On the right side, I've plugged the red variable first and

I've stacked the blue variable on top of that.

On the right it's easy to see that the red variable isn't changing

across the horizontal axis.

On the left, you can still see that the red variable isn't changing, but

you have to see that because the lengths aren't changing.

Whereas on the right, the actual positions aren't changing.

Which further verifies that position is a stronger indicator than the length of

the actual data values in terms of mapping two geometric and

spacial display elements.

You can avoid that stacking order problem by using diverging stacked bar charts if

you have the red variable creating bars growing upwards, and

the blue variable creating bars growing downward.

But that stops working when you have three or more variables.

Another way to avoid having one variable dominate the other in stacking order

is to stack them and then re-center the bar across the horizontal axis.

So instead of growing up from the horizontal axis

the bar is just centered at the horizontal axis.

And this give us the same displacement on the top and

the bottom of the bar chart, but it helps us to pay more

attention to the areas of the elements instead of their actual position.

It's much better if you have continuously changing horizontal and vertical

axis variables to use stacked line graphs, then stacked bar charts.

If we can connect their data values by lines instead of having them

discrete bars, then it's easier to see when areas are changing

versus areas are remaining constant as they move across even

though they're displaced by the variance of other variables.

[MUSIC]