Update app.py

app.py

@@ -1,199 +1,243 @@
-import gradio as gr
-import pandas as pd
-import numpy as np
-import joblib
-import pickle
-
-# --- 1. LOAD MODEL & DATA ---
-model = joblib.load('xgboost_valorant_model.pkl')
-
-with open('player_knowledge_base.pkl', 'rb') as f:
-    knowledge_base = pickle.load(f)
-
-with open('training_columns.pkl', 'rb') as f:
-    training_columns = pickle.load(f)
-
-# Daftar Map & Agent (Sesuaikan dengan data Anda)
-MAP_LIST = ['Ascent', 'Bind', 'Breeze', 'Fracture', 'Haven', 'Icebox', 'Lotus', 'Pearl', 'Split', 'Sunset', 'Abyss']
-AGENT_LIST = ['Jett', 'Raze', 'Reyna', 'Phoenix', 'Yoru', 'Neon', 'Iso', # Duelist
-              'Sova', 'Fade', 'Breach', 'Skye', 'Kayo', 'Gekko',       # Initiator
-              'Omen', 'Brimstone', 'Viper', 'Astra', 'Harbor', 'Clove', 'Waylay', # Controller
-              'Sage', 'Cypher', 'Killjoy', 'Chamber', 'Deadlock', 'Vyse', 'Tejo', 'Veto'] # Sentinel
-
-# --- 2. FUNGSI PREDIKSI ---
-def predict_match(map_name, t1_side, 
-                  t1_p1_name, t1_p1_agent, t1_p2_name, t1_p2_agent, 
-                  t1_p3_name, t1_p3_agent, t1_p4_name, t1_p4_agent, t1_p5_name, t1_p5_agent,
-                  t2_p1_name, t2_p1_agent, t2_p2_name, t2_p2_agent, 
-                  t2_p3_name, t2_p3_agent, t2_p4_name, t2_p4_agent, t2_p5_name, t2_p5_agent):
-    
-    # A. Buat DataFrame Input Kosong
-    input_data = pd.DataFrame(columns=training_columns)
-    input_data.loc[0] = 0 # Isi baris pertama dengan 0 semua
-    
-    # B. Helper: Ambil Stats dari Knowledge Base
-    def get_stats(player, agent):
-        # Jika player/agent ada di database, ambil statsnya
-        # Jika tidak (player baru), pakai nilai default (WR=0.5, Exp=0)
-        key = (player, agent)
-        stats = knowledge_base.get(key, {'Agent_WR': 0.5, 'General_WR': 0.5, 'Exp': 0})
-        return stats
-
-    # C. Isi Fitur Winrate (Skenario 4 Logic)
-    # Team 1
-    t1_inputs = [
-        (t1_p1_name, t1_p1_agent), (t1_p2_name, t1_p2_agent), (t1_p3_name, t1_p3_agent),
-        (t1_p4_name, t1_p4_agent), (t1_p5_name, t1_p5_agent)
-    ]
-    
-    t1_general_wrs = []
-    
-    for i, (p_name, p_agent) in enumerate(t1_inputs):
-        stats = get_stats(p_name, p_agent)
-        idx = i + 1
-        
-        # Isi ke DataFrame (Cek apakah kolom ada di training_columns agar tidak error)
-        if f'T1_P{idx}_Agent_WR' in training_columns:
-            input_data.at[0, f'T1_P{idx}_Agent_WR'] = stats['Agent_WR']
-        if f'T1_P{idx}_General_WR' in training_columns:
-            input_data.at[0, f'T1_P{idx}_General_WR'] = stats['General_WR']
-        if f'T1_P{idx}_Agent_Exp' in training_columns:
-            input_data.at[0, f'T1_P{idx}_Agent_Exp'] = stats['Exp']
-            
-        t1_general_wrs.append(stats['General_WR'])
-
-    # Team 2 (Lakukan hal sama)
-    t2_inputs = [
-        (t2_p1_name, t2_p1_agent), (t2_p2_name, t2_p2_agent), (t2_p3_name, t2_p3_agent),
-        (t2_p4_name, t2_p4_agent), (t2_p5_name, t2_p5_agent)
-    ]
-    t2_general_wrs = []
-    
-    for i, (p_name, p_agent) in enumerate(t2_inputs):
-        stats = get_stats(p_name, p_agent)
-        idx = i + 1
-        if f'T2_P{idx}_Agent_WR' in training_columns:
-            input_data.at[0, f'T2_P{idx}_Agent_WR'] = stats['Agent_WR']
-        if f'T2_P{idx}_General_WR' in training_columns:
-            input_data.at[0, f'T2_P{idx}_General_WR'] = stats['General_WR']
-        if f'T2_P{idx}_Agent_Exp' in training_columns:
-            input_data.at[0, f'T2_P{idx}_Agent_Exp'] = stats['Exp']
-        t2_general_wrs.append(stats['General_WR'])
-
-    # D. Hitung WR Diff
-    if 'WR_Diff' in training_columns:
-        t1_avg = np.mean(t1_general_wrs)
-        t2_avg = np.mean(t2_general_wrs)
-        input_data.at[0, 'WR_Diff'] = t1_avg - t2_avg
-
-    # E. One-Hot Encoding Manual (Set Nilai 1)
-    # Map
-    map_col = f'MAP_{map_name}'
-    if map_col in training_columns:
-        input_data.at[0, map_col] = 1
-        
-    # Start Side (Jika ada fitur ini)
-    if 'T1_StartSide_Defense' in training_columns:
-        input_data.at[0, 'T1_StartSide_Defense'] = 1 if t1_side == 'Defense' else 0
-    if 'T2_StartSide_Defense' in training_columns: # Biasanya kebalikan T1
-        input_data.at[0, 'T2_StartSide_Defense'] = 1 if t1_side == 'Attack' else 0
-
-    # Agent OHE
-    # T1 Agents
-    for _, agent in t1_inputs:
-        # Jika pakai logika "T1_P1_Agent_Jett" (Slot Based)
-        # Atau "T1_Has_Jett" (Aggregated Based) - Sesuaikan dengan model terakhir Anda!
-        
-        # Asumsi: Masih pakai Slot Based (sesuai kode terakhir di chat)
-        # Kita loop kolom training untuk mencocokkan
-        pass 
-        
-    # Karena OHE manual agak ribet mencocokkan nama kolom tepatnya, 
-    # kita gunakan cara pintar: Loop semua input agent, cari kolom yang sesuai di training_columns
-    
-    # Set T1 Agents
-    for i, (_, agent) in enumerate(t1_inputs):
-        col_name = f'T1_P{i+1}_Agent_{agent}'
-        if col_name in training_columns:
-            input_data.at[0, col_name] = 1
-            
-    # Set T2 Agents
-    for i, (_, agent) in enumerate(t2_inputs):
-        col_name = f'T2_P{i+1}_Agent_{agent}'
-        if col_name in training_columns:
-            input_data.at[0, col_name] = 1
-
-    # --- 3. PREDIKSI ---
-    # Convert ke float agar aman
-    input_data = input_data.astype(float)
-    
-    prob = model.predict_proba(input_data)[0]
-    win_prob_t1 = prob[1] # Probabilitas kelas 1 (T1 Win)
-    
-    winner = "TEAM 1" if win_prob_t1 > 0.5 else "TEAM 2"
-    confidence = win_prob_t1 if win_prob_t1 > 0.5 else 1 - win_prob_t1
-    
-    return f"{winner} WIN ({confidence:.1%} Confidence)"
-
-# --- 3. UI GRADIO ---
-with gr.Blocks(title="Valorant Match Predictor") as demo:
-    gr.Markdown("# 🔮 Valorant AI Predictor (Skenario 4)")
-    gr.Markdown("Prediksi pemenang berdasarkan sejarah performa player dan komposisi agent.")
-    
-    with gr.Row():
-        with gr.Column():
-            map_input = gr.Dropdown(MAP_LIST, label="Select Map")
-            side_input = gr.Radio(["Attack", "Defense"], label="Team 1 Start Side", value="Attack")
-    
-    with gr.Row():
-        # Team 1 Inputs
-        with gr.Column():
-            gr.Markdown("### 🔵 Team 1 Roster")
-            t1_p1_n = gr.Textbox(label="P1 Name", placeholder="e.g. f0rsakeN")
-            t1_p1_a = gr.Dropdown(AGENT_LIST, label="P1 Agent")
-            
-            t1_p2_n = gr.Textbox(label="P2 Name", placeholder="e.g. mindfreak")
-            t1_p2_a = gr.Dropdown(AGENT_LIST, label="P2 Agent")
-            
-            t1_p3_n = gr.Textbox(label="P3 Name")
-            t1_p3_a = gr.Dropdown(AGENT_LIST, label="P3 Agent")
-            
-            t1_p4_n = gr.Textbox(label="P4 Name")
-            t1_p4_a = gr.Dropdown(AGENT_LIST, label="P4 Agent")
-            
-            t1_p5_n = gr.Textbox(label="P5 Name")
-            t1_p5_a = gr.Dropdown(AGENT_LIST, label="P5 Agent")
-
-        # Team 2 Inputs
-        with gr.Column():
-            gr.Markdown("### 🔴 Team 2 Roster")
-            t2_p1_n = gr.Textbox(label="P1 Name", placeholder="e.g. Tenz")
-            t2_p1_a = gr.Dropdown(AGENT_LIST, label="P1 Agent")
-            
-            t2_p2_n = gr.Textbox(label="P2 Name", placeholder="e.g. Zekken")
-            t2_p2_a = gr.Dropdown(AGENT_LIST, label="P2 Agent")
-            
-            t2_p3_n = gr.Textbox(label="P3 Name")
-            t2_p3_a = gr.Dropdown(AGENT_LIST, label="P3 Agent")
-            
-            t2_p4_n = gr.Textbox(label="P4 Name")
-            t2_p4_a = gr.Dropdown(AGENT_LIST, label="P4 Agent")
-            
-            t2_p5_n = gr.Textbox(label="P5 Name")
-            t2_p5_a = gr.Dropdown(AGENT_LIST, label="P5 Agent")
-
-    btn = gr.Button("🚀 PREDICT WINNER", variant="primary")
-    output = gr.Label(label="Prediction Result")
-    
-    # Connect Button
-    btn.click(
-        fn=predict_match,
-        inputs=[
-            map_input, side_input,
-            t1_p1_n, t1_p1_a, t1_p2_n, t1_p2_a, t1_p3_n, t1_p3_a, t1_p4_n, t1_p4_a, t1_p5_n, t1_p5_a,
-            t2_p1_n, t2_p1_a, t2_p2_n, t2_p2_a, t2_p3_n, t2_p3_a, t2_p4_n, t2_p4_a, t2_p5_n, t2_p5_a
-        ],
-        outputs=output
-    )
-
+import gradio as gr
+import pandas as pd
+import numpy as np
+import joblib
+import pickle
+
+# --- 1. LOAD MODEL & DATA ---
+# Pastikan file-file .pkl ini ada di folder yang sama saat di-upload ke Hugging Face
+try:
+    model = joblib.load('xgboost_valorant_model.pkl')
+    with open('player_knowledge_base.pkl', 'rb') as f:
+        knowledge_base = pickle.load(f)
+    with open('training_columns.pkl', 'rb') as f:
+        training_columns = pickle.load(f)
+    print("✅ Model dan Data berhasil dimuat.")
+except Exception as e:
+    print(f"❌ Error loading files: {e}")
+    # Dummy data untuk mencegah crash saat development lokal tanpa file
+    knowledge_base = {}
+    training_columns = []
+
+# Daftar Map & Agent (Sesuaikan dengan data Anda)
+MAP_LIST = ['Ascent', 'Bind', 'Breeze', 'Fracture', 'Haven', 'Icebox', 'Lotus', 'Pearl', 'Split', 'Sunset', 'Abyss']
+AGENT_LIST = ['Jett', 'Raze', 'Reyna', 'Phoenix', 'Yoru', 'Neon', 'Iso', # Duelist
+              'Sova', 'Fade', 'Breach', 'Skye', 'Kayo', 'Gekko',       # Initiator
+              'Omen', 'Brimstone', 'Viper', 'Astra', 'Harbor', 'Clove', 'Waylay', # Controller
+              'Sage', 'Cypher', 'Killjoy', 'Chamber', 'Deadlock', 'Vyse', 'Tejo', 'Veto'] # Sentinel
+
+# --- 2. FUNGSI PREDIKSI ---
+def predict_match(map_name, t1_side, 
+                  t1_p1_name, t1_p1_agent, t1_p2_name, t1_p2_agent, 
+                  t1_p3_name, t1_p3_agent, t1_p4_name, t1_p4_agent, t1_p5_name, t1_p5_agent,
+                  t2_p1_name, t2_p1_agent, t2_p2_name, t2_p2_agent, 
+                  t2_p3_name, t2_p3_agent, t2_p4_name, t2_p4_agent, t2_p5_name, t2_p5_agent):
+    
+    # A. Buat DataFrame Input Kosong
+    input_data = pd.DataFrame(columns=training_columns)
+    input_data.loc[0] = 0 # Isi baris pertama dengan 0 semua
+    
+    # B. Helper: Ambil Stats dari Knowledge Base
+    def get_stats(player, agent):
+        # Normalisasi input nama player (strip whitespace)
+        player_clean = player.strip() if player else ""
+        key = (player_clean, agent)
+        
+        # Ambil stats, default 0.5 jika tidak ada
+        stats = knowledge_base.get(key, {'Agent_WR': 0.5, 'General_WR': 0.5, 'Exp': 0})
+        return stats
+
+    # C. Isi Fitur Winrate (Skenario 4 Logic)
+    # Team 1 Inputs
+    t1_inputs = [
+        (t1_p1_name, t1_p1_agent), (t1_p2_name, t1_p2_agent), (t1_p3_name, t1_p3_agent),
+        (t1_p4_name, t1_p4_agent), (t1_p5_name, t1_p5_agent)
+    ]
+    t1_general_wrs = []
+    
+    for i, (p_name, p_agent) in enumerate(t1_inputs):
+        stats = get_stats(p_name, p_agent)
+        idx = i + 1
+        
+        if f'T1_P{idx}_Agent_WR' in training_columns:
+            input_data.at[0, f'T1_P{idx}_Agent_WR'] = stats['Agent_WR']
+        if f'T1_P{idx}_General_WR' in training_columns:
+            input_data.at[0, f'T1_P{idx}_General_WR'] = stats['General_WR']
+        if f'T1_P{idx}_Agent_Exp' in training_columns:
+            input_data.at[0, f'T1_P{idx}_Agent_Exp'] = stats['Exp']
+            
+        t1_general_wrs.append(stats['General_WR'])
+
+    # Team 2 Inputs
+    t2_inputs = [
+        (t2_p1_name, t2_p1_agent), (t2_p2_name, t2_p2_agent), (t2_p3_name, t2_p3_agent),
+        (t2_p4_name, t2_p4_agent), (t2_p5_name, t2_p5_agent)
+    ]
+    t2_general_wrs = []
+    
+    for i, (p_name, p_agent) in enumerate(t2_inputs):
+        stats = get_stats(p_name, p_agent)
+        idx = i + 1
+        
+        if f'T2_P{idx}_Agent_WR' in training_columns:
+            input_data.at[0, f'T2_P{idx}_Agent_WR'] = stats['Agent_WR']
+        if f'T2_P{idx}_General_WR' in training_columns:
+            input_data.at[0, f'T2_P{idx}_General_WR'] = stats['General_WR']
+        if f'T2_P{idx}_Agent_Exp' in training_columns:
+            input_data.at[0, f'T2_P{idx}_Agent_Exp'] = stats['Exp']
+            
+        t2_general_wrs.append(stats['General_WR'])
+
+    # D. Hitung WR Diff
+    if 'WR_Diff' in training_columns:
+        t1_avg = np.mean(t1_general_wrs)
+        t2_avg = np.mean(t2_general_wrs)
+        input_data.at[0, 'WR_Diff'] = t1_avg - t2_avg
+
+    # E. One-Hot Encoding Manual
+    # Map
+    map_col = f'MAP_{map_name}'
+    if map_col in training_columns:
+        input_data.at[0, map_col] = 1
+        
+    # Start Side
+    if 'T1_StartSide_Defense' in training_columns:
+        input_data.at[0, 'T1_StartSide_Defense'] = 1 if t1_side == 'Defense' else 0
+    if 'T2_StartSide_Defense' in training_columns:
+        input_data.at[0, 'T2_StartSide_Defense'] = 1 if t1_side == 'Attack' else 0
+
+    # Agent OHE (Slot Based)
+    # T1 Agents
+    for i, (_, agent) in enumerate(t1_inputs):
+        col_name = f'T1_P{i+1}_Agent_{agent}'
+        if col_name in training_columns:
+            input_data.at[0, col_name] = 1
+            
+    # T2 Agents
+    for i, (_, agent) in enumerate(t2_inputs):
+        col_name = f'T2_P{i+1}_Agent_{agent}'
+        if col_name in training_columns:
+            input_data.at[0, col_name] = 1
+
+    # ==============================================================================
+    # --- BAGIAN DEBUGGING YANG DITAMBAHKAN ---
+    # Log ini akan muncul di Terminal / Logs Hugging Face, bukan di UI
+    print("\n" + "="*40)
+    print(f"🔍 DEBUGGING PREDICTION INPUT")
+    print(f"Map: {map_name} | T1 Side: {t1_side}")
+    print("-" * 40)
+    
+    print("🔵 TIM 1 STATS:")
+    for i, (p, a) in enumerate(t1_inputs):
+        # Ambil nilai yang sudah masuk ke input_data
+        wr = input_data.at[0, f'T1_P{i+1}_Agent_WR']
+        gen_wr = input_data.at[0, f'T1_P{i+1}_General_WR']
+        exp = input_data.at[0, f'T1_P{i+1}_Agent_Exp']
+        
+        status = "✅ FOUND" if wr != 0.5 or gen_wr != 0.5 else "❌ NOT FOUND (Using Default)"
+        print(f"   P{i+1}: {p:<15} ({a:<8}) | WR: {wr:.2f} | GenWR: {gen_wr:.2f} | Exp: {exp:<3} -> {status}")
+
+    print("-" * 40)
+    print("🔴 TIM 2 STATS:")
+    for i, (p, a) in enumerate(t2_inputs):
+        wr = input_data.at[0, f'T2_P{i+1}_Agent_WR']
+        gen_wr = input_data.at[0, f'T2_P{i+1}_General_WR']
+        exp = input_data.at[0, f'T2_P{i+1}_Agent_Exp']
+        
+        status = "✅ FOUND" if wr != 0.5 or gen_wr != 0.5 else "❌ NOT FOUND (Using Default)"
+        print(f"   P{i+1}: {p:<15} ({a:<8}) | WR: {wr:.2f} | GenWR: {gen_wr:.2f} | Exp: {exp:<3} -> {status}")
+
+    if 'WR_Diff' in training_columns:
+        print("-" * 40)
+        print(f"📊 Final WR_Diff (T1 - T2): {input_data.at[0, 'WR_Diff']:.4f}")
+        if input_data.at[0, 'WR_Diff'] > 0:
+            print("   -> Tim 1 Secara Statistik Lebih Unggul.")
+        elif input_data.at[0, 'WR_Diff'] < 0:
+            print("   -> Tim 2 Secara Statistik Lebih Unggul.")
+        else:
+            print("   -> Kekuatan Statistik Seimbang (Netral).")
+            
+    print("="*40 + "\n")
+    # ==============================================================================
+
+    # --- 3. PREDIKSI ---
+    input_data = input_data.astype(float)
+    
+    try:
+        prob = model.predict_proba(input_data)[0]
+        win_prob_t1 = prob[1]
+        
+        if win_prob_t1 > 0.5:
+            winner = "🔵 TEAM 1 WINS"
+            confidence = win_prob_t1
+            color = "blue"
+        else:
+            winner = "🔴 TEAM 2 WINS"
+            confidence = 1 - win_prob_t1
+            color = "red"
+            
+        # Format Output Debugging ke UI juga (Optional)
+        debug_msg = "✅ Data found for players."
+        # Cek jika banyak player not found (WR_Diff 0)
+        if 'WR_Diff' in input_data and input_data.at[0, 'WR_Diff'] == 0:
+             debug_msg = "⚠️ WARNING: Player stats not found (WR_Diff = 0). Check spelling!"
+
+        return f"{winner}\nConfidence: {confidence:.1%}\n({debug_msg})"
+        
+    except Exception as e:
+        return f"Error during prediction: {str(e)}"
+
+# --- 3. UI GRADIO ---
+with gr.Blocks(title="Valorant Match Predictor") as demo:
+    gr.Markdown("# 🔮 Valorant AI Predictor (Skenario 4)")
+    gr.Markdown("Prediksi pemenang berdasarkan sejarah performa player dan komposisi agent.")
+    
+    with gr.Row():
+        with gr.Column():
+            map_input = gr.Dropdown(MAP_LIST, label="Select Map")
+            side_input = gr.Radio(["Attack", "Defense"], label="Team 1 Start Side", value="Attack")
+    
+    with gr.Row():
+        # Team 1 Inputs
+        with gr.Column():
+            gr.Markdown("### 🔵 Team 1 Roster")
+            t1_p1_n = gr.Textbox(label="P1 Name", placeholder="e.g. f0rsakeN")
+            t1_p1_a = gr.Dropdown(AGENT_LIST, label="P1 Agent")
+            t1_p2_n = gr.Textbox(label="P2 Name", placeholder="e.g. mindfreak")
+            t1_p2_a = gr.Dropdown(AGENT_LIST, label="P2 Agent")
+            t1_p3_n = gr.Textbox(label="P3 Name")
+            t1_p3_a = gr.Dropdown(AGENT_LIST, label="P3 Agent")
+            t1_p4_n = gr.Textbox(label="P4 Name")
+            t1_p4_a = gr.Dropdown(AGENT_LIST, label="P4 Agent")
+            t1_p5_n = gr.Textbox(label="P5 Name")
+            t1_p5_a = gr.Dropdown(AGENT_LIST, label="P5 Agent")
+
+        # Team 2 Inputs
+        with gr.Column():
+            gr.Markdown("### 🔴 Team 2 Roster")
+            t2_p1_n = gr.Textbox(label="P1 Name", placeholder="e.g. Tenz")
+            t2_p1_a = gr.Dropdown(AGENT_LIST, label="P1 Agent")
+            t2_p2_n = gr.Textbox(label="P2 Name", placeholder="e.g. Zekken")
+            t2_p2_a = gr.Dropdown(AGENT_LIST, label="P2 Agent")
+            t2_p3_n = gr.Textbox(label="P3 Name")
+            t2_p3_a = gr.Dropdown(AGENT_LIST, label="P3 Agent")
+            t2_p4_n = gr.Textbox(label="P4 Name")
+            t2_p4_a = gr.Dropdown(AGENT_LIST, label="P4 Agent")
+            t2_p5_n = gr.Textbox(label="P5 Name")
+            t2_p5_a = gr.Dropdown(AGENT_LIST, label="P5 Agent")
+
+    btn = gr.Button("🚀 PREDICT WINNER", variant="primary")
+    output = gr.Label(label="Prediction Result")
+    
+    # Connect Button
+    btn.click(
+        fn=predict_match,
+        inputs=[
+            map_input, side_input,
+            t1_p1_n, t1_p1_a, t1_p2_n, t1_p2_a, t1_p3_n, t1_p3_a, t1_p4_n, t1_p4_a, t1_p5_n, t1_p5_a,
+            t2_p1_n, t2_p1_a, t2_p2_n, t2_p2_a, t2_p3_n, t2_p3_a, t2_p4_n, t2_p4_a, t2_p5_n, t2_p5_a
+        ],
+        outputs=output
+    )
+
 demo.launch()