Spaces:
Sleeping
Sleeping
File size: 4,669 Bytes
b099c5d |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 |
import numpy as np
import pandas as pd
from lightgbm import LGBMRegressor
from sklearn.multioutput import MultiOutputRegressor
from utils.metrics import mae, bias
from pathlib import Path
# ---------------------------------------------------------
# CONFIG
# ---------------------------------------------------------
HORIZON = 14 # FreshRetailNet setup: 14-step forecast
TRAIN_DATA_PATH = Path("data/processed/lgbm_ready/train.csv")
TARGET_DATA_PATH = Path("data/processed/lgbm_ready/target.csv")
METRICS_PATH = Path("metrics/lgbm_metrics.csv")
PREDICTIONS_PATH = Path("metrics/lgbm_predictions.csv")
INFERENCE_PATH = Path("data/processed/lgbm_ready/inference/inference_train.csv")
VALIDATION_PATH = Path("data/processed/lgbm_ready/inference/inference_target.csv")
# ---------------------------------------------------------
# Base single-target LGBM
# ---------------------------------------------------------
base_lgbm = LGBMRegressor(
n_estimators=500,
learning_rate=0.05,
max_depth=-1,
subsample=0.8,
colsample_bytree=0.8,
random_state=42,
n_jobs=-1,
verbose=0,
)
model = MultiOutputRegressor(base_lgbm)
def build_features(df: pd.DataFrame) -> pd.DataFrame:
"""
For LGBM multi-output here, each row is:
id | t1 | t2 | ... | tN
We drop 'id' and use the time columns as features/targets.
"""
return df.drop(columns=["id"])
# ---------------------------------------------------------
# Load data
# ---------------------------------------------------------
train_data = pd.read_csv(TRAIN_DATA_PATH)
target_data = pd.read_csv(TARGET_DATA_PATH)
inference_data = pd.read_csv(INFERENCE_PATH)
validation_data = pd.read_csv(VALIDATION_PATH)
# Sanity
# print(train_data.head())
# print(target_data.head())
# print(inference_data.head())
# print(validation_data.head())
train_df = build_features(train_data)
target_df = build_features(target_data)
inference_df = build_features(inference_data)
validation_df = build_features(validation_data)
# ---------------------------------------------------------
# Fit model
# ---------------------------------------------------------
model.fit(train_df, target_df)
# Predict for inference windows
preds = model.predict(inference_df)
# If you want integer-unit forecasts, uncomment:
# preds = np.round(preds)
# ---------------------------------------------------------
# Build long-format prediction DataFrame
# ---------------------------------------------------------
preds_df = pd.DataFrame(preds) # shape: [n_rows, HORIZON]
# Long form: one row per (row_index, horizon_step)
preds_df_long = preds_df.stack().reset_index()
preds_df_long.columns = ["id_index", "h_raw", "forecast"]
# Map back to actual ids using validation_data row order
preds_df_long["id"] = validation_data.iloc[preds_df_long["id_index"]]["id"].values
preds_df_long["h"] = preds_df_long["h_raw"].astype(int) + 1 # 1..HORIZON
preds_df_long["model"] = "lightgbm"
preds_df_final = preds_df_long[["id", "model", "h", "forecast"]]
print(preds_df_final.head(HORIZON))
# ---------------------------------------------------------
# Build long-format truth for validation horizon
# ---------------------------------------------------------
long = validation_data.melt(
id_vars=["id"],
var_name="h",
value_name="sales",
)
long["h"] = long["h"].astype(int)
# ---------------------------------------------------------
# Compute metrics per id
# ---------------------------------------------------------
metrics_rows = []
for (sku_id, model_name), g_fore in preds_df_final.groupby(["id", "model"]):
g_test = long[long["id"] == sku_id].copy()
if g_test["h"].max() < HORIZON:
# safety check – skip if somehow shorter
continue
merged = pd.merge(
g_test[["id", "h", "sales"]],
g_fore[["id", "h", "forecast"]],
on=["id", "h"],
how="inner",
)
if merged.empty:
continue
y_true = merged["sales"].values
y_pred = merged["forecast"].values
m = mae(y_true, y_pred)
b = bias(y_true, y_pred)
s = m + abs(b)
metrics_rows.append(
{
"id": sku_id,
"model": model_name,
"mae": float(m),
"bias": float(b),
"score": float(s),
}
)
metrics_df = pd.DataFrame(metrics_rows)
metrics_df.to_csv(METRICS_PATH, index=False)
preds_df_final.to_csv(PREDICTIONS_PATH, index=False)
print("Saved:")
print(f" - forecasts → {PREDICTIONS_PATH}")
print(f" - metrics → {METRICS_PATH}")
|