3D Scatterplot: Multi Value Plot, coloring by column

Question

I want to perform a 3d scatterplot with a dataframe, which has the following format:

df = pd.DataFrame({"Date": ['2021-01-01', '2021-01-02', '2021-01-03', '2021-01-04'],
           "A_x1": [1, 2, 2, 2],
           "A_x2": [9, 2, 2, 3],
           "A_x3": [1, 3, 2, 9],
           "B_x1": [1, 8, 2, 3],
           "B_x2": [3, 8, 9, 3],
           "B_x3": [2, 4, 5, 5],
           "C_x1": [2, 6, 5, 2],
           "C_x2": [4, 8, 1, 3],
           "C_x3": [6, 9, 5, 7]})

Date	A_x1	A_x2	A_x3	B_x1	B_x2	B_x3	C_x1	C_x2	C_x3	D_x1
2021-01-01	1	9	1	1	3	2	2	4	6	...
2021-01-02	2	2	3	8	8	4	6	8	9	...
2021-01-03	2	2	2	2	9	5	5	1	5	...
2021-01-04	2	3	9	3	3	5	2	3	7	...

As you could guess: The 3 axis of the 3d Scatterplot shall be x1, x2 and x3. So I have 3 variables for 3 axis, but multiple values for each row. I want to plot the values of A_x1/2/3, B_x1/2/3 etc. to the respective point and color them (f. ex. A = red, B = green, C = blue etc.).

I tried to use matplotlib and plotly but I'm open to any other libraries. To get an dataframe or array for all x_1 values I use the following code.

df_x_1 = df.filter(like='1') #df x_1
x_1 = df_x_1.to_numpy() #arr_x_1

This is the simpliest scatterplot in plotly, works fine:

import plotly.express as px
fig = px.scatter_3d(df, 
                    x='A_x1', 
                    y='A_x2', 
                    z='A_x3',
                    #color='species'
                    )
fig.show()

Part of the problem, which has been solved by @Ynjxsjmh spoilered:

But this obv. plots the x1, x2, x3 values for A (=3 columns), I want all >!columns to be included. I want to do something like this but I get different errors. Tried with >!dataframe and arrays. code

fig = px.scatter_3d(x=df.filter(like='1').values.ravel('F'),
                    y=df.filter(like='2').values.ravel('F'),
                    z=df.filter(like='3').values.ravel('F'),
                    color = ( df.filter(like='3').values.ravel('F')*df.filter(like='2').values.ravel('F')*df.filter(like='1').values.ravel('F') )**(1/3)

                    )
fig.show()

This code works now. The datapoints (f. ex. A_x1,x2,x3 are presentet at the corrects spots). What toping is still unclear: Coloring.

Now I'm coloring the datapoints according to their geometrical size by doing color=(x_1x_2x_3)^(1/3)

What i want: Color the Datapoints according to the name of column or the first row of dataframe or whatever (I will have to add this row, but that shall not be a problem).

Any ideas? Thank you!

Answer 1

x, y and z of plotly.express.scatter_3d() should be str or int or Series or array-like. df.filter(like='1') returns a dataframe.

You can use numpy.ravel() to flatten the values in column direction.

fig = px.scatter_3d(x=df.filter(like='1').values.ravel('F'),
                    y=df.filter(like='2').values.ravel('F'),
                    z=df.filter(like='3').values.ravel('F'),
                    )