Without Ground Truth. Open Source. For Data Scientists. Used by Data Scientists at Get started today Use it the way you like: In a Jupyter Notebook. In a CLI. In Docker.
$ pip install nannyml
$ conda install -c conda-forge nannyml
Is not knowing your model performance causing you sleepless nights? What nannyML does for you Your Monitoring Flow NannyML empowers you with the ability to estimate the performance of your deployed machine learning models. It is completely model-agnostic and currently supports all tabular use cases, classification and regression. (NLP and CV work with a bit of hacking ;) )
Know the business impact of your models Define a cost/benefit matrix Set a custom threshold Get alerted when expected business value drops Estimate Business Impact Focus on a single performance metric Estimate model performance of classification models with CBPE
roc_auc, f1, precision, recall, specificity, accuracy or any of the confusion matrix metrics Estimate model performance of regression models with DLE
MAE, MAPE, MSE, RMSE, MSLE, RMSLE Faster root cause analysis Detect changes in your data as a whole through multivariate feature drift PCA based Data Reconstruction Detect changes in individual features Detect changes in your target distribution and model output
Kolmogorov-Smirnov Test, Jensen-Shannon Distance, Wasserstein Distance, Hellinger Distance, Chi-squared Test, L-Infinity Distance Leverage Covariate Shift Detection You're in good company 200+
Monthly Active Instances
1000's
of models monitored by NannyML
27,000+
Data Scientists across LinkedIn and GitHub