事先我已经跑好数据，产出了一个包含
    R：Recency，最近一次消费距今天数
    F：最近一周订单
    M：最近一周用户消费额
    三个数据的表。

    1. import pandas as pd
    2. df = pd.read_excel('/Users/zhaoningbo/dataset/rfm.xlsx')

    image.png

    应该做一个异常值检验，这次先跳过。

    1. import plotly.plotly as py
    2. import pandas as pd
    4. scatter = dict(
    5.     mode = "markers",
    6.     name = "y",
    7.     type = "scatter3d",    
    8.     x = df['r'], y = df['f'], z = df['m'],
    9.     marker = dict( size=2, color="rgb(23, 190, 207)" )
    10. )
    11. layout = dict(
    12.     title = '3d point clustering',
    13.     scene = dict(
    14.         xaxis = dict( zeroline=False ),
    15.         yaxis = dict( zeroline=False ),
    16.         zaxis = dict( zeroline=False ),
    17.     )
    18. )
    19. fig = dict( data=[scatter], layout=layout )
    20. # Use py.iplot() for IPython notebook
    21. plotly.offline.init_notebook_mode() #初始化jupyter notebook中的绘图模式
    22. plotly.offline.iplot(fig, filename='3d point clustering')

    image.png

    1. #转换数据格式
    2. tmp=np.array([df.r,df.f,df.m]).T
    3. #调用python关于机器学习sklearn库中的KMeans
    4. from sklearn.cluster import KMeans
    5. #设置分为3类，并训练数据
    6. kms=KMeans(n_clusters=3)
    7. y=kms.fit_predict(tmp)
    8. #将分类结果以散点图形式展示
    9. y

    array([0, 2, 1, …, 2, 1, 0], dtype=int32)

    1. df["type"] = y
    2. d0 = df[df.type==0]
    3. d1 = df[df.type==1]
    4. d2 = df[df.type==2]
    5. df.type.value_counts()

    0 587
    2 264
    1 227
    Name: type, dtype: int64

    1. import plotly.plotly as py
    2. import pandas as pd
    4. scatter0 = dict(
    5.     mode = "markers",
    6.     name = "d0",
    7.     type = "scatter3d",    
    8.     x = d0['r'], y = d0['f'], z = d0['m'],
    9.     marker = dict( size=2, color="rgb(23, 190, 207)" )
    10. )
    12. scatter1 = dict(
    13.     mode = "markers",
    14.     name = "d1",
    15.     type = "scatter3d",    
    16.     x = d1['r'], y = d1['f'], z = d1['m'],
    17.     marker = dict( size=2, color="rgb(49,54,149)" )
    18. )
    20. scatter2 = dict(
    21.     mode = "markers",
    22.     name = "d2",
    23.     type = "scatter3d",    
    24.     x = d2['r'], y = d2['f'], z = d2['m'],
    25.     marker = dict( size=2, color="rgb(253,174,97)" )
    26. )
    28. fig = dict( data=[scatter0,scatter1,scatter2], layout=layout )
    29. # Use py.iplot() for IPython notebook
    30. plotly.offline.init_notebook_mode() #初始化jupyter notebook中的绘图模式
    31. plotly.offline.iplot(fig, filename='3d point clustering')

    image.png

    因为一开始就选择了用户价值模型，所以没有做降维。