CGMissingData 0.1.7

MICE + ARIMA + XGBoost to handle missing values of CGM device

CGMissingData

CGMissingData is a simple missing-data benchmarking package that runs:

MICE imputation (IterativeImputer)

ARIMA

XGBoost

Your CSV must include at least these columns:

Glucose value (target) = glucose_col

TimeSeries = time series data

Timestamp data= timestamp_col

subject ID = subjectid

How to Run?

1. Envirionment Setup- cd "C:\Path\To\Your\Project"

Create a virtual environment python -m venv .venv

Activate the environment ..venv\Scripts\activate

2. Install python python -m pip install --upgrade pip pip install -e .

3. ..venv\Scripts\python.exe -c "from CGMmissingData import run_missingness_benchmark; r=run_missingness_benchmark('MyData.csv', mask_rates=[0.05, 0.10, 0.20, 0.30, 0.40]); print(r); r.to_csv('results.csv', index=False)" #Ensure your dataset (e.g., MyData.csv) is located in the project file. Execute the benchmark directly from the CLI to generate a results.csv file:

Using Google Colab?

1. !pip -q install CGMissingData==0.1.2 (change the version number depending our new release. You can also try with !pip -q install CGMissingData)

2. from CGMissingData import run_missingness_benchmark

3. df = "/content/drive/MyDrive/CGMExampleData.csv" # your dataset path

4. results = run_missingness_benchmark( "CGMExampleData.csv", # or df mask_rates=[0.05, 0.10, 0.20, 0.30, 0.40] )

5. print(results)

6. results.to_csv("results.csv", index=False)

cgmmissing

CGM missing value imputation pipeline:

1. Convert timestamp to a numeric equal-interval TimeSeries using CGManalyzer::equalInterval.fn via rpy2
2. Add lag features per subjectid: lag1, lag2, lag3, rollmean
3. Compute missing rate on glucose_value
4. If missing rate <= 5%: MICE + ARIMA (segmentwise over missing blocks) Else: MICE + XGBoost
5. Output original dataset plus imputed_glucose_value

Install

pip install -e .