adj_matrix.py

def get_adj_matrix(df, ml_format = True):

  """where df should be one hot of labels"""

  all_categoris = list(df.columns)
  label_freq    = {}

  for i in all_categoris:
    label_freq[i] = df[i].value_counts()[1]
  
  u     = np.diag(np.ones(df.shape[1], dtype=bool))
  adj_m = df.T.dot(df) * (~u)
  adj_m = adj_m.to_numpy()

  if ml_format:
    data = {'adj': adj_m, 'nums': np.array(list(label_freq.values()))}
    return data
  else:
    return adj_m