Notes-in-Machine-Learning-with-Python (4)

发表于 | 更新于 | 分类于

Clustering and Unsupervised Machine Learning

dealing with non-numeric data

  • get the unique value of the columns
  • set the numbers of value id

  • use the id as the numeric data of this column

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    def handle_non_numerical_data(df):
          columns = df.columns.values
          for column in columns:
              text_digit_vals = {}
              def convert_to_int(val):
                  return text_digit_vals[val]

              if df[column].dtype != np.int64 and df[column].dtype != np.float64:
                  column_contents = df[column].values.tolist()
                  unique_elements = set(column_contents)
                  x = 0
                  for unique in unique_elements:
                      if unique not in text_digit_vals:
                          text_digit_vals[unique] = x
                          x+=1

                  df[column] = list(map(convert_to_int, df[column]))

          return df

  • set() —> get the unique values

  • map() —> apply the function for element in the list

K - Means

  • X = preprocessing.scale(X) is necessarily!

  • df.drop(['ticket'], 1, inplace = True) Do not forget the Parameters

    the ‘1’ and ‘inplace = True’

  • np.linalg.norm() —> calculate the distance between two points, there is many options , use it well

  • np.average —> axis = 0 means A 2-dimensional array has two corresponding axes: the first running vertically downwards across rows (axis 0), and the second running horizontally across columns (axis 1).

  • An 3d figure see mplot3d

    1
    2
    3
    4
    5
    6
    7
    import matplotlib.pyplot as plt
    from mpl_toolkits.mplot3d import Axes3D

    ...

    fig = plt.figure()
    ax = fig.add_subplot(111, projection='3d')

Mean - Shift

  • initialize one point as centroid, set bandwidth and class order

  • find all points in the circle (centroid, bandwidth), and set the possibility of class(point) += 1

  • calculate the vector of all the points in class, shift = sum(vectors)

  • center = center + shift

  • replay the 2-4 step, make a convergence mean - shift

  • if the centroid is in the circle of an existed centroid, merge the two class, otherwise make a new class

  • replay 1-5 until all the points are marked

  • classify: find index of max(class(point)) set index as the class
  • print pd.df.describe() —> show more simply one