पायथन का उपयोग करके पावर बीआई कॉहोर्ट विश्लेषण दोहराएं

%%time
#   
path_to_data = "C:/Users/Pavel/Documents/Demo/"
# 
df = pd.read_csv(os.path.join(path_to_data, "ohortAnalysis_2016_2018.csv"), sep=";", parse_dates=["date"], dayfirst=True)

%%time
#  ,      
df.sort_values(["user_id","date"], inplace = True)

%%time
#       user_id. 
df_groupby_user = df.groupby(by = ["user_id"]).agg({"date": "min", "amount": ["sum","count"]})
df_groupby_user.reset_index(inplace = True)
# 
new_columns = ["user_id","first_date_transaction", "total_amount_user","count_transaction_user"]
df_groupby_user.columns = new_columns

%%time
#  -
df_groupby_user["first_transaction"] = df_groupby_user["first_date_transaction"].dt.strftime('%Y-%m')

%%time
# 
df_final = pd.merge(df, df_groupby_user, how = "left", on = "user_id")

%%time
#   "    "
df_final["delta_days"] = df_final["date"] - df_final["first_date_transaction"]
df_final["delta_days"] = df_final["delta_days"].apply(lambda x: x.days)

%%time
#  . 
cut_labels_days = [x for x in range (30, 1230, 30)]
cut_bins_days = [x for x in range (0, 1230, 30)]
df_final["cohort_days"] = pd.cut(df_final["delta_days"], bins = cut_bins_days, labels=cut_labels_days, right = True)
%%time
#     .   fillna   !
df_final["cohort_days"] = df_final["cohort_days"].cat.add_categories([0])
df_final["cohort_days"].fillna(0, inplace = True)
%%time
#     .   fillna   !
df_final["cohort_days"] = df_final["cohort_days"].cat.add_categories([0])
df_final["cohort_days"].fillna(0, inplace = True)
#    .    ,  "0"        ,     
#    .
df_final["cohort_days"] = df_final["cohort_days"].astype(int)

%%time
#  
df_pivot_table = pd.pivot_table(df_final, values=["amount"], index=["first_transaction"], columns=["cohort_days"], aggfunc=np.sum, fill_value = 0)

%%time
#     amount
df_pivot_table_cumsum = df_final.groupby(by = ["first_transaction","cohort_days"]).agg({"amount": ["sum"]}).groupby(level=0).cumsum().reset_index()
df_pivot_table_cumsum.columns = ["first_transaction","cohort_days","cumsum_amount"]
%%time
#     
df_pivot_table_cumsum_chart = copy.deepcopy(df_pivot_table_cumsum)
#     ,       Y.
df_pivot_table_cumsum_chart["cumsum_amount"]=round(df_pivot_table_cumsum_chart["cumsum_amount"]/1000000, 2)

%%time
df_pivot_table_cumsum_chart.pivot(index="cohort_days", columns="first_transaction", values="cumsum_amount").plot(figsize = (15,11))