added an overview with links and leaderboard

app.py

@@ -8,10 +8,12 @@ from components import (
     logos,
     model_selector,
     header,
+    overview_view,
 )
 import utils
 
 PAGES = [
+    "Overview",
     "Buildings",
     "Models",
     "Performance",
@@ -42,8 +44,21 @@ models = sorted(data["model"].unique().tolist())
 with st.sidebar:
     logos()
     view = st.selectbox("View", PAGES, index=0)
-    models_to_plot = model_selector(models)
-    st.subheader("Refresh data")
+
+    if view == "Performance" or view == "Computational Resources":
+        models_to_plot = model_selector(models)
+
+    if view == "Overview":
+        st.header("Sources")
+        st.link_button("GitHub Repository", url="https://github.com/attila-balint-kul/energy-forecast-benchmark-toolkit", use_container_width=True)
+        st.link_button("Documentation", url="https://attila-balint-kul.github.io/energy-forecast-benchmark-toolkit/", use_container_width=True)
+        st.link_button("Electricity Demand Dataset", url="https://huggingface.co/datasets/EDS-lab/electricity-demand", use_container_width=True)
+        st.link_button("HuggingFace Organization", url="https://huggingface.co/EDS-lab", use_container_width=True)
+
+        st.header("Other Dashboards")
+        st.link_button("PV Generation", url="https://huggingface.co/spaces/EDS-lab/EnFoBench-PVGeneration", use_container_width=True)
+
+    st.header("Refresh data")
     refresh = st.button(
         "Refresh", use_container_width=True, help="Fetch the latest data from W&B"
     )
@@ -54,7 +69,9 @@ with st.sidebar:
 
 header()
 
-if view == "Buildings":
+if view == "Overview":
+    overview_view(data)
+elif view == "Buildings":
     buildings_view(data)
 elif view == "Models":
     models_view(data)

components.py

@@ -2,6 +2,8 @@ import pandas as pd
 import streamlit as st
 import plotly.express as px
 
+from utils import get_leaderboard
+
 
 def header() -> None:
     st.title("EnFoBench - Electricity Demand")
@@ -51,6 +53,42 @@ def model_selector(models: list[str]) -> set[str]:
     return models_to_plot
 
 
+def overview_view(data):
+    st.markdown("""
+    [EnFoBench](https://github.com/attila-balint-kul/energy-forecast-benchmark-toolkit) 
+    is a community driven benchmarking framework for energy forecasting models. 
+    
+    This dashboard presents the results of the electricity demand forecasting usecase. All models were cross-validated
+    on **365 days** of day ahead forecasting horizon *(10AM until midnight of the next day)*.
+    """)
+
+    st.divider()
+    st.markdown("## Leaderboard")
+
+    leaderboard = get_leaderboard(data, ["MAE.mean", "RMSE.mean", "rMAE.mean"])
+
+    left, middle, right = st.columns(3)
+    with left:
+        best_models_mae = leaderboard.sort_values("MAE.mean", ascending=False).head(10).sort_values("MAE.mean")
+        fig = px.bar(best_models_mae, x="MAE.mean",  y=best_models_mae.index)
+        fig.update_layout(title="Top 10 models by MAE", xaxis_title="", yaxis_title="Model")
+        st.plotly_chart(fig, use_container_width=True)
+
+    with middle:
+        best_models_mae = leaderboard.sort_values("RMSE.mean", ascending=False).head(10).sort_values("RMSE.mean")
+        fig = px.bar(best_models_mae, x="RMSE.mean",  y=best_models_mae.index)
+        fig.update_layout(title="Top 10 models by RMSE", xaxis_title="", yaxis_title="")
+        st.plotly_chart(fig, use_container_width=True)
+
+    with right:
+        best_models_mae = leaderboard.sort_values("rMAE.mean", ascending=False).head(10).sort_values("rMAE.mean")
+        fig = px.bar(best_models_mae, x="rMAE.mean",  y=best_models_mae.index)
+        fig.update_layout(title="Top 10 models by rMAE", xaxis_title="", yaxis_title="")
+        st.plotly_chart(fig, use_container_width=True)
+
+    st.dataframe(leaderboard, use_container_width=True)
+
+
 def buildings_view(data):
     buildings = (
         data[

utils.py

@@ -27,3 +27,18 @@ def get_wandb_data(entity: str, project: str, api_key: str, job_type: str) -> pd
     runs_df = pd.concat([summary_df, config_df], axis=1)
     runs_df.index = name_list
     return runs_df
+
+
+def get_leaderboard(runs_df: pd.DataFrame, metrics: list[str]) -> pd.DataFrame:
+    leaderboard = pd.DataFrame(
+        index=runs_df['model'].unique(),
+        columns=metrics
+    ).fillna(0)
+
+    for _, building_df in runs_df.groupby("unique_id"):
+        for column in leaderboard.columns:
+            best_model = building_df.loc[building_df[column].idxmin()].model
+            leaderboard.loc[best_model, column] += 1
+
+    leaderboard = leaderboard.sort_values(by=list(leaderboard.columns), ascending=False)
+    return leaderboard