Spaces:
Runtime error
Runtime error
File size: 4,221 Bytes
6eff894 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 | import os
import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import Pipeline
from sklearn.metrics import (
confusion_matrix, classification_report, roc_auc_score,
precision_recall_fscore_support
)
df = pd.read_csv("results/hourly.csv", parse_dates=["time"])
H = 6
precip_next = np.zeros(len(df), dtype=int)
prec = df["precip_mm"].values
for i in range(len(prec) - H):
precip_next[i] = 1 if np.any(prec[i+1:i+1+H] > 0) else 0
df = df.iloc[:len(precip_next) - (0)].copy()
df["rain_next6h"] = precip_next[:len(df)]
features = [
"temp_c","humidity","cloudcover","pressure","wind_speed",
"precip_mm","rain_mm"
]
X = df[features].values
y = df["rain_next6h"].values
print("Class balance (0=no-rain, 1=rain-in-next6h):", np.bincount(y))
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.3, shuffle=False
)
clf = Pipeline([
("scaler", StandardScaler()),
("logreg", LogisticRegression(max_iter=500, class_weight="balanced"))
])
clf.fit(X_train, y_train)
proba = clf.predict_proba(X_test)[:, 1]
pred_050 = (proba >= 0.50).astype(int)
cm = confusion_matrix(y_test, pred_050)
print("\n📊 Confusion Matrix (thr=0.50)")
print(cm)
prec, rec, f1, _ = precision_recall_fscore_support(
y_test, pred_050, average="binary", zero_division=0
)
try:
auc = roc_auc_score(y_test, proba)
except ValueError:
auc = float("nan")
print(f"Precision: {prec:.3f} Recall: {rec:.3f} F1: {f1:.3f} ROC-AUC: {auc:.3f}")
print("\nDetailed report:")
print(classification_report(y_test, pred_050, digits=3, zero_division=0))
# Baselines
always_no = np.zeros_like(y_test)
p0, r0, f10, _ = precision_recall_fscore_support(
y_test, always_no, average="binary", zero_division=0
)
print("\n🧠 Baseline — always 'no rain'")
print(f"Precision: {p0:.3f} Recall: {r0:.3f} F1: {f10:.3f}")
# Persistence baseline
recent_rain = (
pd.Series(df["precip_mm"])
.rolling(window=H, min_periods=1)
.sum()
.shift(1)
.fillna(0)
> 0
).astype(int).values
prev6_test = recent_rain[-len(y_test):]
pp, rp, f1p, _ = precision_recall_fscore_support(y_test, prev6_test, average="binary", zero_division=0)
print("\n🧠 Baseline — persistence (prev 6h)")
print(f"Precision: {pp:.3f} Recall: {rp:.3f} F1: {f1p:.3f}")
# Threshold tuning
thr_recall = 0.35
thr_precision = 0.65
pred_recall = (proba >= thr_recall).astype(int)
pred_precision = (proba >= thr_precision).astype(int)
pr_recall, rc_recall, f1_recall, _ = precision_recall_fscore_support(
y_test, pred_recall, average="binary", zero_division=0
)
pr_precision, rc_precision, f1_precision, _ = precision_recall_fscore_support(
y_test, pred_precision, average="binary", zero_division=0
)
print(f"\n🎛️ Threshold {thr_recall:.2f} → Precision: {pr_recall:.3f} Recall: {rc_recall:.3f} F1: {f1_recall:.3f}")
print(f"🎛️ Threshold {thr_precision:.2f} → Precision: {pr_precision:.3f} Recall: {rc_precision:.3f} F1: {f1_precision:.3f}")
import joblib
os.makedirs("models", exist_ok=True)
joblib.dump(clf, "models/rain_classifier_hourly.joblib")
print("\n💾 Saved: models/rain_classifier_hourly.joblib")
meta = {
"horizon_hours": H,
"features": features,
"thresholds": {
"default": 0.50,
"high_recall": thr_recall,
"high_precision": thr_precision,
},
"metrics": {
"default": {"precision": float(prec), "recall": float(rec), "f1": float(f1)},
"high_recall": {
"precision": float(pr_recall),
"recall": float(rc_recall),
"f1": float(f1_recall),
},
"high_precision": {
"precision": float(pr_precision),
"recall": float(rc_precision),
"f1": float(f1_precision),
},
"baseline_persistence": {
"precision": float(pp),
"recall": float(rp),
"f1": float(f1p),
},
},
}
with open("models/rain_model_meta.json", "w") as fh:
import json
json.dump(meta, fh, indent=2)
print("📝 Saved: models/rain_model_meta.json")
|