Update app.py

app.py

@@ -14,7 +14,7 @@ import matplotlib.pyplot as plt
 from matplotlib.ticker import FuncFormatter
 
 import plotly.graph_objects as go
-from sklearn.metrics import mean_squared_error, mean_absolute_error
+from sklearn.metrics import mean_squared_error  # MAPE implemented manually
 
 # =========================
 # Constants / Config
@@ -145,7 +145,20 @@ add_password_gate()
 # =========================
 # Utilities
 # =========================
-def rmse(y_true, y_pred): return float(np.sqrt(mean_squared_error(y_true, y_pred)))
+def rmse(y_true, y_pred): 
+    return float(np.sqrt(mean_squared_error(y_true, y_pred)))
+
+def mape(y_true, y_pred, eps: float = 1e-8) -> float:
+    """
+    Mean Absolute Percentage Error in PERCENT.
+    Rows where |actual| < eps are ignored to avoid division issues.
+    """
+    a = np.asarray(y_true, dtype=float)
+    p = np.asarray(y_pred, dtype=float)
+    denom = np.where(np.abs(a) < eps, np.nan, np.abs(a))
+    pct = np.abs(a - p) / denom * 100.0
+    val = np.nanmean(pct)
+    return float(val) if np.isfinite(val) else float("nan")
 
 def render_bo_bd_note():
     st.markdown(
@@ -368,7 +381,7 @@ def _x_range_for_tracks(df, cols):
     x_series = pd.concat([pd.to_numeric(df[c], errors="coerce") for c in cols if c in df], ignore_index=True)
     x_lo, x_hi = float(x_series.min()), float(x_series.max())
     pad = 0.03 * (x_hi - x_lo if x_hi > x_lo else 1.0)
-    xmin, xmax = x_lo - pad, x_hi + pad
+    xmin, xmax = x_lo - x_pad, x_hi + x_pad = x_lo - pad, x_hi + pad  # keep variable names consistent
     tick0 = _nice_tick0(xmin, step=max((xmax - xmin)/10.0, 0.1))
     return xmin, xmax, tick0
 
@@ -660,10 +673,10 @@ if st.session_state.app_step == "dev":
         te[PRED_BD] = model_bd.predict(Xte_bd)
 
         st.session_state.results["Train"]=tr; st.session_state.results["Test"]=te
-        st.session_state.results["m_train_bo"]={"R": pearson_r(tr[TARGET_BO], tr[PRED_BO]), "RMSE": rmse(tr[TARGET_BO], tr[PRED_BO]), "MAE": mean_absolute_error(tr[TARGET_BO], tr[PRED_BO])}
-        st.session_state.results["m_train_bd"]={"R": pearson_r(tr[TARGET_BD], tr[PRED_BD]), "RMSE": rmse(tr[TARGET_BD], tr[PRED_BD]), "MAE": mean_absolute_error(tr[TARGET_BD], tr[PRED_BD])}
-        st.session_state.results["m_test_bo"] ={"R": pearson_r(te[TARGET_BO], te[PRED_BO]), "RMSE": rmse(te[TARGET_BO], te[PRED_BO]), "MAE": mean_absolute_error(te[TARGET_BO], te[PRED_BO])}
-        st.session_state.results["m_test_bd"] ={"R": pearson_r(te[TARGET_BD], te[PRED_BD]), "RMSE": rmse(te[TARGET_BD], te[PRED_BD]), "MAE": mean_absolute_error(te[TARGET_BD], te[PRED_BD])}
+        st.session_state.results["m_train_bo"]={"R": pearson_r(tr[TARGET_BO], tr[PRED_BO]), "RMSE": rmse(tr[TARGET_BO], tr[PRED_BO]), "MAPE": mape(tr[TARGET_BO], tr[PRED_BO])}
+        st.session_state.results["m_train_bd"]={"R": pearson_r(tr[TARGET_BD], tr[PRED_BD]), "RMSE": rmse(tr[TARGET_BD], tr[PRED_BD]), "MAPE": mape(tr[TARGET_BD], tr[PRED_BD])}
+        st.session_state.results["m_test_bo"] ={"R": pearson_r(te[TARGET_BO], te[PRED_BO]), "RMSE": rmse(te[TARGET_BO], te[PRED_BO]), "MAPE": mape(te[TARGET_BO], te[PRED_BO])}
+        st.session_state.results["m_test_bd"] ={"R": pearson_r(te[TARGET_BD], te[PRED_BD]), "RMSE": rmse(te[TARGET_BD], te[PRED_BD]), "MAPE": mape(te[TARGET_BD], te[PRED_BD])}
 
         tr_min = tr[st.session_state["FEATURES"]].min().to_dict(); tr_max = tr[st.session_state["FEATURES"]].max().to_dict()
         st.session_state.train_ranges = {f:(float(tr_min[f]), float(tr_max[f])) for f in st.session_state["FEATURES"]}
@@ -674,11 +687,11 @@ if st.session_state.app_step == "dev":
         c1, c2, c3 = st.columns(3)
         c1.metric(f"R ({name})",     f"{m['R']:.3f}")
         c2.metric(f"RMSE ({name})",  f"{m['RMSE']:.2f}")
-        c3.metric(f"MAE ({name})",   f"{m['MAE']:.2f}")
+        c3.metric(f"MAPE (%) ({name})",   f"{m['MAPE']:.2f}%")
 
     def _dev_block(df, mbo, mbd):
         _metrics_block("BO", mbo); _metrics_block("BD", mbd)
-        st.markdown("<div style='text-align:left;font-size:0.8em;color:#6b7280;margin-top:-16px;margin-bottom:8px;'><strong>R</strong> = Pearson correlation • <strong>RMSE/MAE</strong> in MW (pcf)</div>", unsafe_allow_html=True)
+        st.markdown("<div style='text-align:left;font-size:0.8em;color:#6b7280;margin-top:-16px;margin-bottom:8px;'><strong>R</strong> = Pearson correlation • <strong>RMSE</strong> in MW (pcf) • <strong>MAPE</strong> in %</div>", unsafe_allow_html=True)
         t1, t2, t3 = st.tabs(["Breakout", "Breakdown", "Combined"])
         with t1:
             left, right = st.columns([3,1], gap="large")
@@ -755,8 +768,8 @@ if st.session_state.app_step == "validate":
                     if pd.api.types.is_numeric_dtype(tbl[c]): tbl[c] = tbl[c].round(2)
                 tbl["Violations"] = pd.DataFrame({f:(df[f]<ranges[f][0])|(df[f]>ranges[f][1]) for f in st.session_state["FEATURES"]}).loc[any_viol].apply(lambda r:", ".join([c for c,v in r.items() if v]), axis=1)
 
-        st.session_state.results["m_val_bo"]={"R": pearson_r(df[TARGET_BO], df[PRED_BO]), "RMSE": rmse(df[TARGET_BO], df[PRED_BO]), "MAE": mean_absolute_error(df[TARGET_BO], df[PRED_BO])}
-        st.session_state.results["m_val_bd"]={"R": pearson_r(df[TARGET_BD], df[PRED_BD]), "RMSE": rmse(df[TARGET_BD], df[PRED_BD]), "MAE": mean_absolute_error(df[TARGET_BD], df[PRED_BD])}
+        st.session_state.results["m_val_bo"]={"R": pearson_r(df[TARGET_BO], df[PRED_BO]), "RMSE": rmse(df[TARGET_BO], df[PRED_BO]), "MAPE": mape(df[TARGET_BO], df[PRED_BO])}
+        st.session_state.results["m_val_bd"]={"R": pearson_r(df[TARGET_BD], df[PRED_BD]), "RMSE": rmse(df[TARGET_BD], df[PRED_BD]), "MAPE": mape(df[TARGET_BD], df[PRED_BD])}
         st.session_state.results["sv_val"]={"n":len(df), "bo_min":float(df[PRED_BO].min()), "bo_max":float(df[PRED_BO].max()),
                                             "bd_min":float(df[PRED_BD].min()), "bd_max":float(df[PRED_BD].max()), "oor":oor_pct}
         st.session_state.results["oor_tbl"]=tbl
@@ -767,12 +780,12 @@ if st.session_state.app_step == "validate":
         c1,c2,c3 = st.columns(3)
         c1.metric("R (Breakout)",     f"{m_bo['R']:.3f}")
         c2.metric("RMSE (Breakout)",  f"{m_bo['RMSE']:.2f}")
-        c3.metric("MAE (Breakout)",   f"{m_bo['MAE']:.2f}")
+        c3.metric("MAPE (%) (Breakout)",   f"{m_bo['MAPE']:.2f}%")
         c1,c2,c3 = st.columns(3)
         c1.metric("R (Breakdown)",     f"{m_bd['R']:.3f}")
         c2.metric("RMSE (Breakdown)",  f"{m_bd['RMSE']:.2f}")
-        c3.metric("MAE (Breakdown)",   f"{m_bd['MAE']:.2f}")
-        st.markdown("<div style='text-align:left;font-size:0.8em;color:#6b7280;margin-top:-16px;margin-bottom:8px;'>R = Pearson correlation</div>", unsafe_allow_html=True)
+        c3.metric("MAPE (%) (Breakdown)",   f"{m_bd['MAPE']:.2f}%")
+        st.markdown("<div style='text-align:left;font-size:0.8em;color:#6b7280;margin-top:-16px;margin-bottom:8px;'>R = Pearson correlation • RMSE in MW (pcf) • MAPE in %</div>", unsafe_allow_html=True)
 
         t1, t2, t3 = st.tabs(["Breakout", "Breakdown", "Combined"])
         with t1: