How the score behaves in each band: count, event rate, the event and non-event distributions, KS, lift, cumulative capture, odds and the score interval of the band, which is what lets a cut-off be read straight from the table. The bands are the deciles of the score on train, applied frozen to the other samples.
Arguments
- x
An object from
scr_scorecard().- sample
NULL(all),"train"or"holdout".
Value
A data.table with one row per sample and band, the band
richest in events first (the riskiest under objective = "risk"):
sample, id, band, n, pct, events, non_events,
event_rate, pct_event and pct_nonevent (the band's share of all
events and of all non-events), woe, min_score, mean_score,
max_score, cum_pct, cum_event_pct, cum_nonevent_pct, ks,
lift, cum_lift, odds and log_odds.
Details
woe is log(pct_event / pct_nonevent), event-oriented like the WOE of
the variables (positive when the band event rate is above the overall
rate) and equal to log_odds in scr_strategy() for the same sample and
bands; when a band has no events or no non-events, 0.5 is added to the
counts of every band for woe only. odds follows the odds
orientation of the scale: non-events per event under higher_is_safer,
events per non-event under higher_is_riskier, with 0.5 added to each
count. log_odds therefore rises with the score under both directions,
and its slope against mean_score can be read against log(2) / pdo.
See also
Other accessors:
scr_funnel(),
scr_gains(),
scr_leakage(),
scr_result,
scr_score_metrics(),
scr_selected()
Examples
cfg <- scr_config(verbose = FALSE, nthread = 1, use_ranger = FALSE,
xgb_rounds = 60, n_boot = 20)
res <- scr_select(scr_demo, "default", config = cfg, drop = "id",
date_col = "ref_date")
sc <- scr_scorecard(res)
scr_score_gains(sc, "holdout")[, .(band, n, event_rate, min_score, max_score, ks)]
#> band n event_rate min_score max_score ks
#> <char> <int> <num> <num> <num> <num>
#> 1: [-Inf,510] 127 0.35433071 452.9460 509.8567 0.1531703
#> 2: (510,524] 128 0.27343750 510.0702 523.3842 0.2478898
#> 3: (524,533] 139 0.26618705 523.6932 533.3423 0.3449428
#> 4: (533,542] 149 0.17449664 533.4699 542.0327 0.3702647
#> 5: (542,550] 151 0.11258278 542.1594 550.2070 0.3420621
#> 6: (550,558] 148 0.12162162 550.3653 557.7466 0.3221272
#> 7: (558,567] 149 0.07382550 557.7815 566.4153 0.2610261
#> 8: (567,577] 128 0.03906250 566.5926 576.4365 0.1828998
#> 9: (577,590] 124 0.03225806 576.6784 590.2077 0.1023536
#> 10: (590, Inf] 157 0.03184713 590.3197 652.3243 0.0000000
scr_score_metrics(sc)
#> Index: <sample>
#> sample direction n events auc auc_lo auc_hi ks
#> <char> <char> <int> <int> <num> <num> <num> <num>
#> 1: train higher_is_safer 2800 399 0.7856428 0.7662170 0.8069608 0.4411466
#> 2: holdout higher_is_safer 1400 203 0.7394060 0.7066284 0.7770357 0.3889033
#> ks_lo ks_hi gini gini_lo gini_hi n_boot level
#> <num> <num> <num> <num> <num> <int> <num>
#> 1: 0.4145334 0.4875854 0.5712856 0.5324341 0.6139216 20 0.95
#> 2: 0.3381932 0.4654704 0.4788120 0.4132569 0.5540713 20 0.95