python plt颜色标签

Question

# x,y,size 데이터 셋팅
x = target_data.accuracy
y =  target_data.f1_score
s = target_data.recall

# 라벨셋팅(순서유의)
users =['dnn', 'random forest', 'extra trees', 'ensemble']

This is a questionable point.

# 컬러셋팅
colors =  list(np.array([0.81520346,0.28735556, 0.6542928, 0.3542928]))
df = pd.DataFrame(dict(accuracy=x, f1_score=y, users=users, s=s, c=colors ))

# 그래프 그리기
ax = df.plot.scatter(x='accuracy', y='f1_score', s=df.s*10,c= df.c,  alpha=0.5)
for i, txt in enumerate(users):
    ax.annotate(txt, (df.accuracy.iat[i],df.f1_score.iat[i]))
plt.show()

I definitely set the color array and matched the graph.

The resulting image is well-formed with graphs and labels, but the color is represented as a black and white image. What's wrong?

Answer 1

Nothing is wrong, but your cmap (Colormap) defaults to a color map where numbers map to opacity. You could change it to whatever you like from this long list: https://matplotlib.org/examples/color/colormaps_reference.html and get a very colorful plot

ax = df.plot.scatter(x='accuracy', y='f1_score', s=df.s*10,c= df.c, 
                 alpha=0.5, cmap='PuOr') # added cmap

Answer 2

If you use just scalar numbers for color setting, it will map to a colormap, which is by default rc image.cmap - and this is grayscale by default. So you can either additionally set another colormap (like eg cmap='viridis' or whatever), or you define your colors as color names, rgb(a) tuples or rgb hex values:

Examples:

colors = ['r', 'g', 'b', 'm']
colors = [[.2, .3, .4], [.7, .3, .4], [.2, .9, .4], [.2, .3, .0]]
colors = ['#a0ff45', '#123456', '#fedcba', '#00ff77']