ts-anomaly-td/ts_anomaly_td/visualize.py

"""可视化模块：双栏对比图（GT 上、TDengine 检测下）。"""

import logging
from pathlib import Path
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
import matplotlib.font_manager as fm

logger = logging.getLogger(__name__)

ALGO_COLORS = {
    "ksigma": "#1f77b4",
    "iqr": "#2ca02c",
    "grubbs": "#9467bd",
    "shesd": "#ff7f0e",
    "lof": "#17becf",
    "sample_ad_model": "#e377c2",
}


def _find_cjk_font():
    """查找可用 CJK 字体。"""
    candidates = ["Noto Serif CJK SC", "Noto Sans CJK SC", "WenQuanYi Micro Hei",
                  "DejaVu Sans", "sans-serif"]
    available = {f.name for f in fm.fontManager.ttflist}
    for name in candidates:
        if name in available:
            return name
    return "sans-serif"


def _compute_iou(gt_windows, det_windows):
    """计算 GT 与检测窗口的 IoU。

    窗口格式: [(start_ts, end_ts), ...]
    IoU = 交集时长 / 并集时长
    """
    if not gt_windows or not det_windows:
        return 0.0

    def overlap(a_start, a_end, b_start, b_end):
        o_start = max(a_start, b_start)
        o_end = min(a_end, b_end)
        return max(0, o_end - o_start)

    intersection = 0.0
    union = 0.0

    for gw in gt_windows:
        union += gw[1] - gw[0]
        for dw in det_windows:
            intersection += overlap(gw[0], gw[1], dw[0], dw[1])

    for dw in det_windows:
        union += dw[1] - dw[0]

    if union == 0:
        return 0.0
    return min(intersection / union, 1.0)


def render_contrast(
    timestamps: list[int],
    values: list[float],
    gt_labels: list[int],
    detection_results: dict,
    output_path: str,
    title: str = "",
) -> str:
    """渲染双栏对比图。

    Args:
        timestamps: 毫秒时间戳列表
        values: 数值列表
        gt_labels: GT 标签（0/1）
        detection_results: {algo: {"windows": [...]}} 来自 detect_all_algos
        output_path: PNG 输出路径
        title: 图表标题

    Returns:
        output_path
    """
    font_name = _find_cjk_font()
    plt.rcParams["font.family"] = font_name
    plt.rcParams["font.size"] = 9

    fig, (ax_gt, ax_det) = plt.subplots(2, 1, sharex=True, figsize=(14, 8))

    t0 = timestamps[0]
    hours = [(t - t0) / 3_600_000 for t in timestamps]

    ax_gt.plot(hours, values, color="black", linewidth=0.6, alpha=0.8)
    ax_gt.set_ylabel("Value")
    ax_gt.set_title(f"Ground Truth — {title}" if title else "Ground Truth")

    in_anomaly = False
    anom_start = None
    gt_windows = []
    for i, label in enumerate(gt_labels):
        if label == 1 and not in_anomaly:
            in_anomaly = True
            anom_start = hours[i]
        elif label == 0 and in_anomaly:
            in_anomaly = False
            ax_gt.axvspan(anom_start, hours[i], alpha=0.25, color="red")
            gt_windows.append((anom_start, hours[i]))

    if in_anomaly:
        ax_gt.axvspan(anom_start, hours[-1], alpha=0.25, color="red")
        gt_windows.append((anom_start, hours[-1]))

    ax_det.plot(hours, values, color="black", linewidth=0.6, alpha=0.3)
    ax_det.set_ylabel("Value")
    ax_det.set_xlabel("Time (hours)")

    all_det_windows = []
    for algo, result in detection_results.items():
        if result.get("error"):
            continue
        color = ALGO_COLORS.get(algo, "#888888")
        for wstart, wend in result.get("windows", []):
            wstart_h = (wstart - timestamps[0]) / 3_600_000 if wstart > timestamps[0] else 0
            wend_h = (wend - timestamps[0]) / 3_600_000
            ax_det.axvspan(wstart_h, wend_h, alpha=0.2, color=color)
            all_det_windows.append((wstart_h, wend_h))

    legend_handles = []
    for algo in detection_results:
        if algo in ALGO_COLORS and not detection_results[algo].get("error"):
            from matplotlib.patches import Patch
            legend_handles.append(Patch(color=ALGO_COLORS[algo], alpha=0.5, label=algo))
    if legend_handles:
        ax_det.legend(handles=legend_handles, loc="upper right", fontsize=7)

    ax_det.set_title(f"TDengine Detection — {title}" if title else "TDengine Detection (6 algos)")

    iou = _compute_iou(gt_windows, all_det_windows)
    total_det = len(all_det_windows)
    stats_text = f"GT windows: {len(gt_windows)}\nDetected: {total_det}\nIoU: {iou:.3f}"
    ax_det.text(
        0.98, 0.97, stats_text,
        transform=ax_det.transAxes,
        verticalalignment="top",
        horizontalalignment="right",
        bbox=dict(boxstyle="round", facecolor="white", alpha=0.8),
        fontsize=8,
    )

    plt.tight_layout()
    Path(output_path).parent.mkdir(parents=True, exist_ok=True)
    fig.savefig(output_path, dpi=150, bbox_inches="tight")
    plt.close(fig)
    logger.info("chart saved: %s", output_path)
    return output_path