finish all analysis and upload data and results

2026-04-26 10:33:02 +09:00
parent 9d76d82f5e
commit fc1386e572
376 changed files with 1059309 additions and 246 deletions
@@ -0,0 +1,92 @@
+import pandas as pd
+import os
+import matplotlib.pyplot as plt
+from sklearn.metrics import mean_squared_error, mean_absolute_error
+
+raw_path  = "results/real_data/raw/dimacs_s42.csv"
+pred_path = "results/real_data/derived/distribution/dimacs_prediction_result/prediction.csv"
+save_folder = "results/real_data/derived/comparison"
+os.makedirs(save_folder, exist_ok=True)
+
+raw  = pd.read_csv(raw_path)
+pred = pd.read_csv(pred_path)
+
+# real: trial 평균
+agg = (
+    raw.groupby("file")
+    .agg(
+        nodes        = ("nodes",        "first"),
+        density      = ("density",      "first"),
+        real_time    = ("time",         "mean"),
+        real_relax   = ("relax_success","mean"),
+    )
+    .reset_index()
+)
+
+# prediction: file 기준으로 merge
+pred_cols = pred[["file", "predict_time", "relax_success"]].rename(
+    columns={"relax_success": "pred_relax"}
+)
+
+merged = agg.merge(pred_cols, on="file", how="inner")
+
+# 오차 계산
+for real_col, pred_col, prefix in [
+    ("real_time",  "predict_time", "time"),
+    ("real_relax", "pred_relax",   "relax"),
+]:
+    merged[f"{prefix}_abs_err"] = (merged[pred_col] - merged[real_col]).abs()
+    merged[f"{prefix}_rel_err"] = merged[f"{prefix}_abs_err"] / merged[real_col]
+
+cols = [
+    "file", "nodes", "density",
+    "real_time",  "predict_time",  "time_abs_err",  "time_rel_err",
+    "real_relax", "pred_relax",    "relax_abs_err", "relax_rel_err",
+]
+merged[cols].to_csv(os.path.join(save_folder, "comparison.csv"), index=False)
+print(merged[cols].to_string(index=False))
+print(f"\nSaved → {save_folder}/comparison.csv")
+
+for real_col, pred_col, label in [
+    ("real_time",  "predict_time", "time"),
+    ("real_relax", "pred_relax",   "relax_success"),
+]:
+    y     = merged[real_col].to_numpy()
+    y_hat = merged[pred_col].to_numpy()
+    ss_res = ((y - y_hat) ** 2).sum()
+    ss_tot = ((y - y.mean()) ** 2).sum()
+    r2   = 1 - ss_res / ss_tot
+    rmse = mean_squared_error(y, y_hat) ** 0.5
+    mae  = mean_absolute_error(y, y_hat)
+    print(f"\n[{label}]  R²={r2:.4f}  RMSE={rmse:.4g}  MAE={mae:.4g}")
+
+
+# ── Plot ───────────────────────────────────────────────────────────────────────
+fig, axes = plt.subplots(1, 2, figsize=(12, 5))
+
+for ax, real_col, pred_col, label in [
+    (axes[0], "real_time",  "predict_time", "Runtime [s]"),
+    (axes[1], "real_relax", "pred_relax",   "Relax Success"),
+]:
+    x = merged[real_col].to_numpy()
+    y = merged[pred_col].to_numpy()
+    mn, mx = min(x.min(), y.min()), max(x.max(), y.max())
+
+    ax.scatter(x, y, s=60, zorder=3)
+    for _, row in merged.iterrows():
+        ax.annotate(
+            row["file"].replace("USA-road-d.", "").replace(".gr", ""),
+            (row[real_col], row[pred_col]),
+            fontsize=6, textcoords="offset points", xytext=(4, 2)
+        )
+    ax.plot([mn, mx], [mn, mx], color="red", linestyle="--", linewidth=1.2, label="y = x")
+    ax.set_xlabel(f"Real {label}")
+    ax.set_ylabel(f"Predicted {label}")
+    ax.set_title(f"Predicted vs Real — {label}")
+    ax.legend(fontsize=8)
+    ax.grid(True, alpha=0.4)
+
+fig.tight_layout()
+fig.savefig(os.path.join(save_folder, "comparison_plot.png"), dpi=300)
+plt.close(fig)
+print(f"Saved → {save_folder}/comparison_plot.png")