add covid code

.github/workflows/submit-covid-forecasts.yaml

@@ -0,0 +1,126 @@
+name: Make COVID Forecast Submission
+
+on:
+  schedule:
+    # Every Wednesday at 6pm CET (17:00 UTC)
+    - cron: '0 17 * * 3'
+  workflow_dispatch:  # Allow manual triggering
+
+jobs:
+  submit-forecast:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+      pull-requests: write
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.10'
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install requests pandas numpy
+
+      - name: Create required directories
+        run: |
+          mkdir -p covid/submissions
+
+      - name: Run covid forecasting script
+        run: python covid/run-covid-forecasts.py
+        env:
+          PYTHONPATH: ${{ github.workspace }}
+
+      - name: Commit forecasts to repository
+        run: |
+          # Configure git
+          git config user.name "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+
+          # Add and commit the new forecasts
+          git add covid/submissions/
+
+          # Only commit if there are changes
+          if git diff --staged --quiet; then
+            echo "No new forecasts to commit"
+          else
+            git commit -m "Store covid forecasts $(date +%Y-%m-%d)"
+            git push
+          fi
+
+      - name: Fork and sync target repository
+        run: |
+          # Install GitHub CLI
+          gh auth login --with-token <<< "${{ secrets.PRIVATE_ACCESS_TOKEN }}"
+
+          # Fork the repository (if not already forked)
+          gh repo fork CDCgov/covid19-forecast-hub --clone=true || true
+
+          # Sync the fork with upstream
+          pushd covid19-forecast-hub
+
+          # Add upstream remote if it doesn't exist
+          git remote add upstream https://github.com/CDCgov/covid19-forecast-hub.git 2>/dev/null || true
+
+          # Set the origin remote with authentication token
+          git remote set-url origin "https://${{ secrets.PRIVATE_ACCESS_TOKEN }}@github.com/${{ github.actor }}/covid19-forecast-hub.git"
+
+          # Fetch and sync with upstream
+          git fetch upstream
+          git checkout main
+          git reset --hard upstream/main
+          git push origin main --force
+
+          popd
+
+      - name: Copy forecast files
+        run: |
+          # Create target directory if it doesn't exist
+          mkdir -p covid19-forecast-hub/model-output/Metaculus-cp
+
+          # Copy new forecasts
+          cp -r covid/submissions/* covid19-forecast-hub/model-output/Metaculus-cp/
+
+      - name: Create Pull Request
+        run: |
+          cd covid19-forecast-hub
+
+          # Setup git config
+          git config user.name "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+
+          # Create a new branch with timestamp to ensure uniqueness
+          BRANCH_NAME="covid-forecast-update-$(date +%Y%m%d-%H%M%S)"
+          git checkout -b $BRANCH_NAME
+
+          # Add and commit changes
+          git add model-output/Metaculus-cp/
+
+          # Only commit if there are changes
+          if git diff --staged --quiet; then
+            echo "No changes to commit"
+            exit 0
+          else
+            git commit -m "Update covid forecasts $(date +%Y-%m-%d)"
+
+            # Set the remote URL with authentication token
+            git remote set-url origin "https://${{ secrets.PRIVATE_ACCESS_TOKEN }}@github.com/${{ github.actor }}/covid19-forecast-hub.git"
+
+            # Force push to fork with the new unique branch
+            git push -f origin $BRANCH_NAME
+
+            # Create PR
+            gh pr create \
+              --title "Update covid forecasts $(date +%Y-%m-%d)" \
+              --body "Automated covid forecast submission from Metaculus" \
+              --repo CDCgov/covid19-forecast-hub \
+              --base main \
+              --head "${{ github.actor }}:$BRANCH_NAME"
+          fi
+        env:
+          PRIVATE_ACCESS_TOKEN: ${{ secrets.PRIVATE_ACCESS_TOKEN }}

covid/Metaculus-cp.yaml

@@ -0,0 +1,28 @@
+team_name: "Metaculus"
+team_abbr: "Metaculus"
+model_name: "Metaculus Community Prediction"
+model_abbr: "cp"
+model_version: "1.0"
+model_contributors:
+    [
+        {
+            "name": "Ryan Beck",
+            "affiliation": "Metaculus",
+            "email": "ryan@metaculus.com",
+        },
+        {
+            "name": "Nikos Bosse",
+            "affiliation": "Metaculus",
+            "email": "nikos@metaculus.com",
+        },
+    ]
+website_url: "https://www.metaculus.com/questions/30049/us-covid-hospitalization-forecasts-2024-25/"
+repo_url: "https://github.com/Metaculus/respiratory-diseases"
+license: "CC-BY-4.0"
+designated_model: true
+team_funding: "This project is supported by the National Science Foundation under Award No. 2438211. Any opinions, findings and conclusions or recommendations expressed in this project are those of Metaculus and our forecasters, and do not necessarily reflect the views of the National Science Foundation."
+methods: "A recency-weighted average of predictions made by forecasters on the Metaculus prediction platform."
+data_inputs: "Users are allowed to make use of any data they choose. The recency-weighted average takes only the numeric forecasts made by forecasters on the platform into account."
+methods_long: "The Metaculus Community Prediction is a consensus of recent forecaster predictions. It is designed to respond to big changes in forecaster opinion while still being fairly insensitive to outliers. For every forecaster, on ly their most recent prediction is kept. Predictions are assigned a number n, from oldest to newest (oldest is 1). Every prediction is weighted proportional to exp(sqrt(n)). Predictions are then aggregated by creating a mixture distribution of all available weighted forecasts."
+ensemble_of_models: true
+ensemble_of_hub_models: false

covid/run-covid-forecasts.py

@@ -0,0 +1,76 @@
+# https://www.metaculus.com/questions/30049/us-covid-hospitalization-forecasts-2024-25/
+
+# The Challenge Period will begin November 20, 2024, and will run until May 31, 2025. Participants are asked to submit weekly nowcasts and forecasts by 11PM Eastern Time each Wednesday (herein referred to as the Forecast Due Date).
+
+import requests
+from datetime import datetime, timedelta
+import pandas as pd
+from utils import internal_to_actual
+import numpy as np
+
+question_id = 30049
+url = f"https://metaculus.com/api/posts/{question_id}"
+response = requests.get(url).json()
+
+# get reference date, which is the saturday following the submission due date
+today = datetime.now().date()  # this is the submission due date, a Tuesday
+days_until_saturday = (5 - today.weekday()) % 7  # 5 is Saturday
+reference_date = today + timedelta(days=days_until_saturday)
+
+forecasts = []
+
+subquestions = response["group_of_questions"]["questions"]
+for subquestion in subquestions:
+    # obtain the target end date
+    question_title = subquestion["title"]
+    target_end_date = question_title.split("(")[1].split(")")[0].strip()
+    target_end_date = datetime.strptime(target_end_date, "%B %d, %Y").date()
+
+    # calculate horizon based on target_end_date and reference_date
+    # target_end_date should be equal to the reference_date + horizon*(7 days).
+    horizon = (target_end_date - reference_date).days // 7
+
+    # # only deal with forecast, if horizon is -1, 0, 1, 2, 3
+    if horizon not in [-1, 0, 1, 2, 3]:
+        continue
+
+    # obtain the scaling of the x-axis
+    range_max = subquestion["scaling"]["range_max"]
+    range_min = subquestion["scaling"]["range_min"]
+    zero_point = subquestion["scaling"]["zero_point"]
+
+    try:
+        cdf = subquestion["aggregations"]["recency_weighted"]["latest"][
+            "forecast_values"
+        ]
+    except TypeError:
+        print(f"No forecast for {question_title}")
+        continue
+
+    internal_x_grid = np.linspace(0, 1, 201)
+    actual_x_grid = internal_to_actual(
+        internal_x_grid, zero_point, range_min, range_max, is_linear=False
+    )
+
+    desired_quantile_levels = np.concatenate(
+        [[0.01, 0.025], np.arange(0.05, 0.95 + 0.05, 0.05), [0.975, 0.99]]
+    ).round(3)
+    desired_quantiles = np.interp(desired_quantile_levels, cdf, actual_x_grid)
+
+    # need a dataframe with the quantiles and the quantile levels
+    latest_forecast_df = pd.DataFrame(
+        {
+            "reference_date": reference_date,
+            "target": "wk inc covid hosp",
+            "horizon": horizon,
+            "target_end_date": target_end_date,
+            "location": "US",
+            "output_type": "quantile",
+            "output_type_id": desired_quantile_levels,
+            "value": desired_quantiles,
+        }
+    )
+    forecasts.append(latest_forecast_df)
+
+forecasts_df = pd.concat(forecasts)
+forecasts_df.to_csv(f"covid/submissions/{reference_date}-Metaculus-cp.csv", index=False)