synthetic_leases.py

now = datetime.now()
start = now.replace(day=1, hour=0, minute=0, second=0, microsecond=0)
start = start + timedelta(days=31)
stop = start + timedelta(days=30 * 36)
 
synth_certs = pd.DataFrame()
for dt in rrule.rrule(rrule.MONTHLY, dtstart=start, until=stop):
   #THE SEASONALITY CURVE IS MEASURED (in SQL) DIRECTLY FROM OUR DATA
   cohort_size = (
       cert_ramp_seasonality_df[
           (cert_ramp_seasonality_df.month_num == dt.month)                    ]
       .fillna(0)
       .iloc[0]["total_new_certs_per_carrier"]
       .astype(int)
   )
   #NEW SALES EXPECTATIONS DRIVEN BY BUSINESS STAKEHOLDERS
   running_new_launched_units += new_units_per_month
   ramp_perc = (
       cert_ramp_seasonality_df[
           (cert_ramp_seasonality_df.month_num == dt.month)
       ]
       .fillna(0)
       .iloc[0]["avg_new_cert_perc"]
   )
   cohort_size += int(ramp_perc * running_new_launched_units)
   #ONCE WE KNOW HOW MANY LEASES WE EXPECT LET'S SAMPLE FROM THE HISTORICAL DATA AND SIMULATE THEIR LIFECYCLES
   cohort = (
       lease_df
       .sample(cohort_size, replace=True)
       .reset_index()
   )
   cohort.move_in_date = dt
  
   surv_samples = trained_survival_curve_for_propert.sample(cohort_size)
   surv_samples["offsets"] = surv_samples.months_into_term.apply(
       lambda x: pd.offsets.DateOffset(months=x)
   )
   cohort.move_out_date = (
       cohort.move_in_date + surv_samples["offsets"]
   )
   #THIS DATAFRAME WILL BECOME OUR “SYNTHETIC LEASE” DIMENSION
   synth_certs = pd.concat([synth_certs] + [cohort])