Update early_warning_module.py

early_warning_module.py

@@ -2,8 +2,8 @@ from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
 from collections import Counter
 from datetime import datetime, timedelta
 import random
+import pandas as pd
 
-# ── Trigger keywords theo từng loại vấn đề ───────────────────────────
 WARNING_RULES = {
     "waste_overflow": {
         "keywords": ["overflow", "overflowing", "full bin", "no bin", "trash everywhere", "garbage everywhere"],
@@ -37,95 +37,71 @@ WARNING_RULES = {
     },
 }
 
-SEVERITY_EMOJI = {
-    "CRITICAL": "🚨",
-    "HIGH":     "🔴",
-    "MEDIUM":   "🟡",
-    "LOW":      "🟢"
-}
+SEVERITY_EMOJI = {"CRITICAL": "🚨", "HIGH": "🔴", "MEDIUM": "🟡", "LOW": "🟢"}
 
-# ── Simulate historical reviews (để demo trend) ───────────────────────
-def simulate_review_history(current_reviews: str, risk_score: float):
-    """Tạo lịch sử 7 ngày để vẽ trend — dựa trên risk score hiện tại."""
-    base = risk_score
-    history = []
+def build_trend_df(risk_score: float) -> pd.DataFrame:
+    """Tạo DataFrame 7 ngày để vẽ line chart thật bằng gr.LinePlot."""
+    rows = []
     for i in range(6, 0, -1):
-        date = (datetime.now() - timedelta(days=i)).strftime("%b %d")
-        # Giả lập fluctuation quanh base
+        date = datetime.now() - timedelta(days=i)
         noise = random.uniform(-0.15, 0.15)
-        score = round(max(0.0, min(1.0, base + noise - 0.05 * i)), 2)
-        history.append({"date": date, "score": score})
-    # Ngày hôm nay = score thật
-    history.append({"date": "Today", "score": round(risk_score, 2)})
-    return history
-
-# ── Main function ─────────────────────────────────────────────────────
-def check_early_warning(reviews_text: str, risk_score: float) -> str:
+        score = round(max(0.0, min(1.0, risk_score + noise - 0.04 * i)), 3)
+        rows.append({"date": date, "risk_score": score, "period": "Historical"})
+    # Hôm nay = điểm thật
+    rows.append({"date": datetime.now(), "risk_score": round(risk_score, 3), "period": "Today"})
+    return pd.DataFrame(rows)
+
+def check_early_warning(reviews_text: str, risk_score: float):
+    """
+    Returns:
+        warning_text (str): markdown output
+        trend_df (pd.DataFrame): dùng cho gr.LinePlot
+    """
+    empty_df = pd.DataFrame({"date": [], "risk_score": []})
+
     if not reviews_text.strip():
-        return "No reviews provided for early warning analysis."
+        return "⚠️ Run Text Analysis first (Tab 1).", empty_df
 
     reviews = [r.strip().lower() for r in reviews_text.strip().split("\n") if r.strip()]
-    analyzer = SentimentIntensityAnalyzer()
 
-    # ── Step 1: Trigger-based detection ──────────────────────────────
+    # ── Trigger detection ─────────────────────────────────────────────
     triggered = []
     for rule_name, rule in WARNING_RULES.items():
         hits = []
         for review in reviews:
             found = [kw for kw in rule["keywords"] if kw in review]
-            if found:
-                hits.extend(found)
+            hits.extend(found)
         if hits:
-            freq = Counter(hits)
             triggered.append({
                 "rule":     rule_name,
                 "message":  rule["message"],
                 "severity": rule["severity"],
-                "hits":     dict(freq),
+                "hits":     dict(Counter(hits)),
                 "count":    len(hits)
             })
 
-    # Sort by severity
     severity_order = {"CRITICAL": 0, "HIGH": 1, "MEDIUM": 2, "LOW": 3}
     triggered.sort(key=lambda x: severity_order.get(x["severity"], 99))
 
-    # ── Step 2: Trend simulation ──────────────────────────────────────
-    history = simulate_review_history(reviews_text, risk_score)
-    trend_values = [h["score"] for h in history]
-
-    # Detect spike: nếu 2 ngày gần nhất tăng > 0.15 = spike
-    spike = False
-    if len(trend_values) >= 3:
-        recent_change = trend_values[-1] - trend_values[-3]
-        spike = recent_change > 0.15
-
-    # Draw trend chart (text-based)
-    trend_rows = ""
-    for h in history:
-        bar_len = int(h["score"] * 20)
-        bar     = "█" * bar_len
-        flag    = " ⚠️" if h["score"] > 0.6 else ""
-        trend_rows += f"| {h['date']} | {bar:<20} | {h['score']:.2f}{flag} |\n"
-
-    # ── Step 3: Format output ─────────────────────────────────────────
+    # ── Trend ─────────────────────────────────────────────���───────────
+    trend_df = build_trend_df(risk_score)
+    trend_values = trend_df["risk_score"].tolist()
+    spike = len(trend_values) >= 3 and (trend_values[-1] - trend_values[-3]) > 0.15
+
+    # ── Format warnings ───────────────────────────────────────────────
     if not triggered:
         warning_section = "✅ **No warning triggers detected.** Reviews appear safe.\n"
     else:
         warning_section = ""
         for t in triggered:
-            emoji = SEVERITY_EMOJI.get(t["severity"], "⚠️")
-            kw_str = ", ".join([f"'{k}' x{v}" for k, v in t["hits"].items()])
+            emoji  = SEVERITY_EMOJI.get(t["severity"], "⚠️")
+            kw_str = ", ".join([f"'{k}' ×{v}" for k, v in t["hits"].items()])
             warning_section += f"""
 #### {emoji} {t["severity"]} — {t["message"]}
 - **Keywords found:** {kw_str}
-- **Frequency:** {t["count"]} mention(s) across reviews
+- **Frequency:** {t["count"]} mention(s)
 """
 
-    spike_alert = ""
-    if spike:
-        spike_alert = "\n> 🚨 **SPIKE DETECTED:** Risk score increased significantly in the last 48 hours — possible emerging issue!\n"
-
-    # Overall alert level
     if any(t["severity"] == "CRITICAL" for t in triggered):
         overall = "🚨 CRITICAL — Immediate intervention required"
     elif any(t["severity"] == "HIGH" for t in triggered):
@@ -135,31 +111,28 @@ def check_early_warning(reviews_text: str, risk_score: float) -> str:
     else:
         overall = "🟢 ALL CLEAR — No immediate warnings"
 
-    output = f"""## ⚡ Early Warning System
+    spike_alert = "\n> 🚨 **SPIKE DETECTED:** Risk increased significantly in last 48h!\n" if spike else ""
+
+    alert_list = "\n".join([
+        f"- **{t['severity']}** → {t['message']}" for t in triggered
+    ]) if triggered else "- No alerts triggered."
+
+    warning_text = f"""## ⚡ Early Warning System
 
 ### Overall Status: {overall}
 {spike_alert}
 
 ---
 
-### 🔍 Warning Triggers Detected
+### 🔍 Warning Triggers
 {warning_section}
 
 ---
 
-### 📈 Risk Score Trend (Last 7 Days)
-
-| Date | Risk Level | Score |
-|------|-----------|-------|
-{trend_rows}
-
-{'⚠️ **Trend is RISING** — investigate before peak season.' if spike else '📉 Trend is stable or improving.'}
-
----
-
 ### 📬 Auto-Alert Summary
-If this were a live system, the following alerts would be sent:
+{alert_list}
 
-{chr(10).join([f"- **{t['severity']}** alert to operator: {t['message']}" for t in triggered]) if triggered else "- No alerts triggered."}
+---
+*(Risk trend chart shown below)*
 """
-    return output
+    return warning_text, trend_df