From d5bac61b47063f92dbe6c2b85d6b0bd54202ad4b Mon Sep 17 00:00:00 2001
From: peterdudfield <peter.dudfield@hotmail.com>
Date: Tue, 12 Dec 2023 15:32:59 +0000
Subject: [PATCH] refactor into metrics file

---
 quartz_solar_forecast/eval/__init__.py |  0
 quartz_solar_forecast/eval/metrics.py  | 29 ++++++++++++++++++++++++++
 quartz_solar_forecast/evaluation.py    |  6 +++---
 tests/eval/test_metrics.py             | 20 ++++++++++++++++++
 4 files changed, 52 insertions(+), 3 deletions(-)
 create mode 100644 quartz_solar_forecast/eval/__init__.py
 create mode 100644 quartz_solar_forecast/eval/metrics.py
 create mode 100644 tests/eval/test_metrics.py

diff --git a/quartz_solar_forecast/eval/__init__.py b/quartz_solar_forecast/eval/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/quartz_solar_forecast/eval/metrics.py b/quartz_solar_forecast/eval/metrics.py
new file mode 100644
index 00000000..94e9ec05
--- /dev/null
+++ b/quartz_solar_forecast/eval/metrics.py
@@ -0,0 +1,29 @@
+def metrics(results_df):
+    """
+    Calculate and print metrics: MAE
+
+    results_df dataframe with the following columns
+    - timestamp
+    - pv_id
+    - horizon_hours
+    - forecast_power
+    - generation_power
+
+    """
+
+    mae = (results_df["forecast_power"] - results_df['generation_power']).abs().mean()
+    print(f"MAE: {mae}")
+
+    # calculate metrics over the different horizons hours
+    # find all unique horizon_hours
+    horizon_hours = results_df["horizon_hours"].unique()
+    for horizon_hour in horizon_hours:
+        # filter results_df to only include the horizon_hour
+        results_df_horizon = results_df[results_df["horizon_hours"] == horizon_hour]
+        mae = (results_df_horizon["forecast_power"] - results_df_horizon['generation_power']).abs().mean()
+        print(f"MAE for horizon {horizon_hour}: {mae}")
+
+    # TODO add more metrics using ocf_ml_metrics
+
+
+
diff --git a/quartz_solar_forecast/evaluation.py b/quartz_solar_forecast/evaluation.py
index a79104b4..8a03a00a 100644
--- a/quartz_solar_forecast/evaluation.py
+++ b/quartz_solar_forecast/evaluation.py
@@ -5,6 +5,7 @@
 This contains 50 sites each with 50 timestamps to make 2500 samples in total.
 
 """
+from quartz_solar_forecast.eval.metrics import metrics
 from quartz_solar_forecast.eval.nwp import get_nwp
 from quartz_solar_forecast.eval.forecast import run_forecast
 from quartz_solar_forecast.eval.utils import combine_forecast_ground_truth
@@ -35,9 +36,8 @@ def run_eval(testset_path):
     results_df.to_csv("results.csv")
 
     # Calculate and print metrics: MAE
-    mae = (results_df["forecast_power"] - results_df['generation_power']).abs().mean()
-    print(f"MAE: {mae}")
-    # TODO: add more metrics using ocf_ml_metrics
+    metrics(results_df)
 
     # Visulisations
 
+
diff --git a/tests/eval/test_metrics.py b/tests/eval/test_metrics.py
new file mode 100644
index 00000000..442cacc6
--- /dev/null
+++ b/tests/eval/test_metrics.py
@@ -0,0 +1,20 @@
+from quartz_solar_forecast.eval.metrics import metrics
+import pandas as pd
+import numpy as np
+
+
+def test_metrics():
+
+    # create a fake dataframe
+
+    results_df = pd.DataFrame(
+        columns=[
+            "id",
+            "timestamp",
+            "horizon_hours",
+            "forecast_power",
+            "generation_power",
+        ], data=np.random.random((100,5)))
+
+    # call the metrics function
+    metrics(results_df)
\ No newline at end of file