Create nodes

nodes.py

def order_brand(PATH_IN):
    ''' List of all brands in each order'''
    # Import DataFrame
    df_rec = pd.read_excel(PATH_IN)

    # Listing Unique Brands
    df_ordbr = pd.DataFrame(df_rec.groupby(['ORDER_NUMBER'])['BRAND'].unique())
    df_ordbr.columns = ['list_brand']
    
    # source = list brands
    list_brand = list(df_rec['BRAND'].unique())

    # boolean column per brand for each order: is brand in order ?
    for br in list_brand:
        df_ordbr[br] = df_ordbr['list_brand'].apply(lambda t: br in t)

    # unique combinations of brands boolean 
    df_con = pd.DataFrame(df_ordbr.reset_index()[df_ordbr.columns[1:]]).drop_duplicates()
    
    return list_brand, df_ordbr, df_con, df_rec

def create_nodes(df_con, n_groups):
    ''' Create nodes from df_con'''
    list_col, list_cont = [], []

    # how many brands are ordered with this brand
    for col in df_con.columns:
        list_col.append(col)
        list_cont.append((df_con[df_con[col] == True].sum() > 0).sum())
    df_nodes = pd.DataFrame({'name': list_col, 'group':list_cont})
    df_nodes.set_index('name', inplace = True)

    # group by range of values
    range_value = np.ceil(df_nodes['group'].max()/n_groups)
    df_nodes['group'] = n_groups - (df_nodes['group']/range_value).apply(np.floor).astype(int)
    
    return df_nodes