预测型标签引擎 — 将未来风险转化为当前可查询标签的供应链预测打标体系

"""
预测型标签引擎 — 供应链供应链预测打标体系
功能：多类型预测标签计算 / 滚动更新 / 置信度管理 / 提前行动映射
输入：SKU库存数据 + 销售历史 + PLT数据
输出：预测标签集 + 置信度 + 行动优先级 + 预测准确率追踪
"""
import numpy as np
import pandas as pd
from datetime import datetime, timedelta
from dataclasses import dataclass, field
from typing import Optional
import warnings
warnings.filterwarnings('ignore')


@dataclass
class PredictiveTag:
    tag_id: str
    horizon_days: int
    value: bool
    confidence: float
    predicted_at: str
    model_name: str
    evidence: dict = field(default_factory=dict)
    action_priority: str = "normal"  # low/normal/high/urgent

    def is_actionable(self, min_confidence: float = 0.75) -> bool:
        return self.value and self.confidence >= min_confidence


class PredictiveTagEngine:
    """预测型标签引擎"""

    def __init__(self, confidence_thresholds: dict = None):
        self.thresholds = confidence_thresholds or {
            "predicted_stockout_7d": 0.75,
            "predicted_stockout_14d": 0.70,
            "predicted_stockout_30d": 0.65,
            "predicted_demand_spike": 0.72,
            "predicted_slow_moving": 0.68,
            "predicted_price_increase": 0.70,
        }
        self.prediction_log = []

    def predict_stockout(self, sku: dict, horizon_days: int,
                          plt_p85: float = 35.0) -> PredictiveTag:
        """
        断货预测：基于DOS + PLT + 需求波动
        DOS = current_inventory / avg_daily_sales
        """
        inventory = sku.get("inventory", 0)
        avg_daily = max(0.1, sku.get("avg_daily_sales_30d", 1.0))
        demand_cv = sku.get("demand_cv", 0.25)
        pending_po = sku.get("pending_po_qty", 0)

        # 有效库存天数（含在途但扣安全库存）
        effective_inventory = inventory + pending_po * 0.8  # 在途80%可期
        dos = effective_inventory / avg_daily

        # 需求上行风险（CV越大，断货风险越高）