# -*- coding: utf-8 -*- """ 华尔街财团投入指数 & 反击情绪指数:从爬虫实时数据计算稳定指标,抑制单条报道导致的剧烈波动。 供 db_merge.merge() 调用,写入同一批 DB 表,前端契约不变。 """ from typing import Optional, Tuple VALUE_CLAMP_MIN = 1 VALUE_CLAMP_MAX = 99 # 华尔街:与上一点平滑,新点权重 WALL_STREET_NEW_WEIGHT = 0.35 # raw 权重;1 - 此值 = 上一点权重,越大曲线越平滑 # 华尔街:两次写入最小间隔(分钟),避免短时间多条报道造成密集锯齿 WALL_STREET_MIN_INTERVAL_MINUTES = 20 # 反击情绪:当前值权重(1 - 此值 = 新 raw 权重) RETALIATION_CURRENT_WEIGHT = 0.8 # 反击情绪:单次更新相对当前值的最大变化幅度(绝对值) RETALIATION_MAX_STEP = 5 def clamp(value: int) -> int: return max(VALUE_CLAMP_MIN, min(VALUE_CLAMP_MAX, int(value))) def smooth_wall_street( raw_value: int, last_value: Optional[int], *, new_weight: float = WALL_STREET_NEW_WEIGHT, ) -> int: """ 华尔街投入指数:用上一点做平滑,避免单条报道 30/80 导致曲线骤变。 若尚无上一点,直接使用限幅后的 raw。 """ raw = clamp(raw_value) if last_value is None: return raw w = 1.0 - new_weight return clamp(round(w * last_value + new_weight * raw)) def wall_street_should_append( last_time_iso: Optional[str], new_time_iso: str, min_interval_minutes: int = WALL_STREET_MIN_INTERVAL_MINUTES, ) -> bool: """ 是否应追加一条华尔街趋势点。若与上一条间隔不足 min_interval_minutes 则跳过, 减少因爬虫短时间多篇报道导致的密集锯齿。 """ if not last_time_iso: return True try: from datetime import datetime last = datetime.fromisoformat(last_time_iso.replace("Z", "+00:00")) new = datetime.fromisoformat(new_time_iso.replace("Z", "+00:00")) delta_min = (new - last).total_seconds() / 60 return delta_min >= min_interval_minutes except Exception: return True def smooth_retaliation( raw_value: int, current_value: int, *, current_weight: float = RETALIATION_CURRENT_WEIGHT, max_step: int = RETALIATION_MAX_STEP, ) -> int: """ 反击情绪指数:先与当前值平滑,再限制单步变化幅度,避免连续多条报道导致快速漂移或抖动。 """ raw = clamp(raw_value) cur = clamp(current_value) smoothed = round(current_weight * cur + (1.0 - current_weight) * raw) smoothed = clamp(smoothed) # 单步变化上限 delta = smoothed - cur if abs(delta) > max_step: step = max_step if delta > 0 else -max_step smoothed = clamp(cur + step) return smoothed