P3-5.2 汇总表如何保留平均值之外的模式¶

# 这个例子在汇总表中保留 early、mid、late 区段模式，避免平均值掩盖它们。
import csv
from collections import Counter
from pathlib import Path

pattern_change_threshold = 0.30
preview_count = 8

data_path = Path("docs/assets/part-03/chapter-05/p3_5_2_segment_patterns.csv")

with data_path.open(newline="", encoding="utf-8") as file:
    summary = []
    for row in csv.DictReader(file):
        numeric = {
            key: float(row[key])
            for key in ["early_flow_mean", "mid_flow_mean", "late_flow_mean"]
        }
        overall_mean = sum(numeric.values()) / len(numeric)
        mid_minus_early = round(numeric["mid_flow_mean"] - numeric["early_flow_mean"], 2)
        late_minus_mid = round(numeric["late_flow_mean"] - numeric["mid_flow_mean"], 2)

        if (
            mid_minus_early > pattern_change_threshold
            and late_minus_mid <= -pattern_change_threshold
        ):
            pattern_note = "mid peak then drop"
        elif (
            abs(mid_minus_early) <= pattern_change_threshold
            and abs(late_minus_mid) <= pattern_change_threshold
        ):
            pattern_note = "flat across segments"
        elif late_minus_mid <= -pattern_change_threshold:
            pattern_note = "late decline after high early/mid"
        else:
            pattern_note = "other segment pattern"

        summary.append(
            {
                **row,
                **numeric,
                "overall_mean": overall_mean,
                "mid_minus_early": mid_minus_early,
                "late_minus_mid": late_minus_mid,
                "pattern_note": pattern_note,
            }
        )

print("1) the same overall mean is not enough")
for row in summary[:preview_count]:
    print(
        f'{row["event_id"]}: overall={row["overall_mean"]:.2f} '
        f'early={row["early_flow_mean"]:.2f} '
        f'mid={row["mid_flow_mean"]:.2f} '
        f'late={row["late_flow_mean"]:.2f}'
    )
print(f"... {len(summary) - preview_count} more event summaries")
print()
print(f"2) pattern counts when threshold = {pattern_change_threshold:.2f}")
for note, count in sorted(Counter(row["pattern_note"] for row in summary).items()):
    print(f"{note}: {count}")
print()
print("3) derived pattern columns for the preview rows")
for row in summary[:preview_count]:
    print(
        f'{row["event_id"]}: '
        f'mid_minus_early={row["mid_minus_early"]:.2f} '
        f'late_minus_mid={row["late_minus_mid"]:.2f} '
        f'-> {row["pattern_note"]}'
    )

1) the same overall mean is not enough
E01: overall=2.40 early=1.80 mid=2.90 late=2.50
E02: overall=2.40 early=2.40 mid=2.40 late=2.40
E03: overall=2.40 early=2.70 mid=2.70 late=1.80
E04: overall=2.40 early=1.90 mid=2.80 late=2.50
E05: overall=2.40 early=2.35 mid=2.45 late=2.40
E06: overall=2.40 early=2.75 mid=2.65 late=1.80
E07: overall=2.40 early=1.70 mid=2.90 late=2.60
E08: overall=2.40 early=2.45 mid=2.35 late=2.40
... 28 more event summaries

2) pattern counts when threshold = 0.30
flat across segments: 12
late decline after high early/mid: 12
mid peak then drop: 12

3) derived pattern columns for the preview rows
E01: mid_minus_early=1.10 late_minus_mid=-0.40 -> mid peak then drop
E02: mid_minus_early=0.00 late_minus_mid=0.00 -> flat across segments
E03: mid_minus_early=0.00 late_minus_mid=-0.90 -> late decline after high early/mid
E04: mid_minus_early=0.90 late_minus_mid=-0.30 -> mid peak then drop
E05: mid_minus_early=0.10 late_minus_mid=-0.05 -> flat across segments
E06: mid_minus_early=-0.10 late_minus_mid=-0.85 -> late decline after high early/mid
E07: mid_minus_early=1.20 late_minus_mid=-0.30 -> mid peak then drop
E08: mid_minus_early=-0.10 late_minus_mid=0.05 -> flat across segments