From 566744766241cfe8664996c448f01edf69681b25 Mon Sep 17 00:00:00 2001 From: itlubber <1830611168@qq.com> Date: Wed, 28 Feb 2024 13:38:25 +0800 Subject: [PATCH] fix rule report --- scorecardpipeline/__init__.py | 2 +- scorecardpipeline/processing.py | 4 +-- scorecardpipeline/rule.py | 56 +++++++++++++++++++++++++++------ 3 files changed, 49 insertions(+), 13 deletions(-) diff --git a/scorecardpipeline/__init__.py b/scorecardpipeline/__init__.py index a017449..a6194ae 100644 --- a/scorecardpipeline/__init__.py +++ b/scorecardpipeline/__init__.py @@ -19,7 +19,7 @@ from .rule import Rule -__version__ = "0.1.30" +__version__ = "0.1.31" __all__ = ( "__version__" , "FeatureSelection", "FeatureImportanceSelector", "StepwiseSelection", "Combiner", "WOETransformer" diff --git a/scorecardpipeline/processing.py b/scorecardpipeline/processing.py index 55fa71f..fc5b045 100644 --- a/scorecardpipeline/processing.py +++ b/scorecardpipeline/processing.py @@ -556,14 +556,14 @@ def export(self, to_json=None): """ return self.combiner.export(to_json=to_json) - def load(self, from_json=None): + def load(self, from_json): """特征分箱器加载离线保存的 json 文件 :param from_json: json 文件的路径 :return: Combiner，特征分箱器 """ - self.combiner.load(from_json=from_json) + self.combiner.load(from_json) return self @classmethod diff --git a/scorecardpipeline/rule.py b/scorecardpipeline/rule.py index 7c7a5a7..30852c6 100644 --- a/scorecardpipeline/rule.py +++ b/scorecardpipeline/rule.py @@ -13,7 +13,7 @@ from sklearn.utils import check_array from sklearn.metrics import f1_score, recall_score, accuracy_score, precision_score -from .processing import feature_bin_stats +from .processing import feature_bin_stats, Combiner def _get_context(X, feature_names): @@ -75,6 +75,24 @@ def json2expr(data, max_index, feature_list): class Rule: def __init__(self, expr): # expr 既可以传递字符串，也可以传递dict + """规则集 + + :param expr: 类似 DataFrame 的 query 方法传参方式即可，目前仅支持数值型变量规则 + + **参考样例** + + >>> from scorecardpipeline import * + >>> target = "creditability" + >>> data = germancredit() + >>> data[target] = data[target].map({"good": 0, "bad": 1}) + >>> data = data.select_dtypes("number") # 暂不支持字符型规则 + >>> rule1 = Rule("duration_in_month < 10") + >>> rule2 = Rule("credit_amount < 500") + >>> rule1.report(data, target=target) + >>> rule2.report(data, target=target) + >>> (rule1 | rule2).report(data, target=target) + >>> (rule1 & rule2).report(data, target=target) + """ self._state = RuleState.INITIALIZED self.expr = expr @@ -113,9 +131,22 @@ def predict(self, X: DataFrame, part=""): # dict预测对应part_dict 、字符 return result - def report(self, datasets, target="target", overdue="overdue", dpd=-1, del_grey=False, valid=None, desc="", return_cols=None, prior_rules=None): + def report(self, datasets, target="target", overdue="overdue", dpd=-1, del_grey=False, desc="", return_cols=None, prior_rules=None) -> pd.DataFrame: + """规则效果报告表格输出 + + :param datasets: 数据集，需要包含 目标变量 或 逾期天数，当不包含目标变量时，会通过逾期天数计算目标变量，同时需要传入逾期定义的DPD天数 + :param target: 目标变量名称，默认 target + :param desc: 规则相关的描述，会出现在返回的表格当中 + :param return_cols: 指定返回的字段列表，默认不传 + :param prior_rules: 先验规则，可以传入先验规则先筛选数据后再评估规则效果 + :param overdue: 逾期天数字读名称 + :param dpd: 逾期定义方式，逾期天数 > DPD 为 1，其他为 0，仅 overdue 字段起作用时有用 + :param del_grey: 是否删除逾期天数 (0, dpd] 的数据，仅 overdue 字段起作用时有用 + + :return: pd.DataFrame，规则效果评估表 + """ if return_cols is None: - return_cols = ['指标名称', "指标含义", '分箱', '样本总数', '样本占比', '好样本数', '好样本占比', '坏样本数', '坏样本占比', '坏样本率', 'LIFT值', '分档KS值'] + return_cols = ['指标名称', "指标含义", '分箱', '样本总数', '样本占比', '好样本数', '好样本占比', '坏样本数', '坏样本占比', '坏样本率', 'LIFT值'] if desc is None or desc == "" and "指标含义" in return_cols: return_cols.remove("指标含义") @@ -130,14 +161,16 @@ def report(self, datasets, target="target", overdue="overdue", dpd=-1, del_grey= rule_expr = self.expr if prior_rules: - prior_tables = prior_rules.report(datasets, target=target, overdue=overdue, dpd=dpd, del_grey=del_grey, valid=valid, desc=desc, return_cols=return_cols, prior_rules=None) + prior_tables = prior_rules.report(datasets, target=target, overdue=overdue, dpd=dpd, del_grey=del_grey, desc=desc, return_cols=return_cols, prior_rules=None) temp = datasets[~prior_rules.predict(datasets)] rule_result = pd.DataFrame({rule_expr: np.where(self.predict(temp), "命中", "未命中"), "target": temp[target].tolist()}) else: prior_tables = pd.DataFrame(columns=return_cols) rule_result = pd.DataFrame({rule_expr: np.where(self.predict(datasets), "命中", "未命中"), "target": datasets[target].tolist()}) - table = feature_bin_stats(rule_result, rule_expr, rules=[["命中"], ["未命中"]], desc=desc, return_cols=return_cols) + combiner = Combiner(target=target) + combiner.load({rule_expr: [["命中"], ["未命中"]]}) + table = feature_bin_stats(rule_result, rule_expr, combiner=combiner, desc=desc, return_cols=return_cols) # 准确率、精确率、召回率、F1分数 metrics = pd.DataFrame({ @@ -149,14 +182,17 @@ def report(self, datasets, target="target", overdue="overdue", dpd=-1, del_grey= }) table = table.merge(metrics, on="分箱", how="left") + # 规则上线后增益评估 + # 坏账率变化情况: 上线后拒绝多少比例的坏客户同时拒绝后坏账水平多少，在原始数据基础上换张改善多少 + total_bad, total = table["坏样本数"].sum(), table["样本总数"].sum() + total_bad_rate = total_bad / total + table["坏账改善"] = (total_bad_rate - (total_bad - table["坏样本数"]) / (total - table["样本总数"])) / total_bad_rate + if prior_rules: prior_tables.insert(loc=0, column="规则分类", value=["先验规则"] * len(prior_tables)) + prior_tables["坏账改善"] = np.nan table.insert(loc=0, column="规则分类", value=["验证规则"] * len(table)) - table = pd.concat([prior_tables, table]) - - # 规则上线后增益评估 - # 通过率变化情况: 上线后拒绝多少客户 - # 坏账率变化情况: 上线后拒绝多少比例的坏客户同时拒绝后坏账水平多少 + table = pd.concat([prior_tables, table]).set_index(["规则分类"]) return table