mindstrong 1.0

Mindstrong Digital Biomarker Model Fitting

This package uses Supervised Kernel Principal Components Analysis with cross validation to fit digital biomarker data to target measurements. The software was written by members of the Mindstrong Health Data Science team:

• Paul Dagum, MD, PhD

• Greg Ryslik, PhD, FCAS, MAAA

• Bob Dougherty, PhD

• Patrick Staples, PhD

Please contact us at [e-mail].

NOTE: If you use this software in your work, please cite the following paper:

Dagum, P. (2018) Digital biomarkers of cognitive function. npj Digital Medicine, issue 1, article 10. DOI: 10.1038/s41746-018-0018-4.

Installation

The easiest way to install the package is via easy_install or pip:

$ pip install mindstrong_biomarker_modelfit

This should also take care of the dependencies (numpy, scipy, pandas, and sklearn).

Usage

Simulated digital biomarker and target measure data are included with the project. To fit a model to these example data:

import numpy as np
import pandas as pd
import os
from mindstrong import mindstrong_modelfit as mindstrong

target_file = mindstrong.get_example_data('example_targets.csv')
feature_file = mindstrong.get_example_data('example_features.csv')
target_colname = 'target1'

# Load target data
target_df = pd.read_csv(target_file)
target_df.set_index('device_id', inplace=True)

# Load Feature Data
feature_df = pd.read_csv(feature_file).set_index(['device_id', 'targetDOY'])

# Cross Validated supervised kernel PCA model-fitting
cvdf, best_model = mindstrong.calculateCrossValidatedCorrelation(target_df,
                                                                 feature_df,
                                                                 target_colname,
                                                                 fold_type='n',
                                                                 n_folds=5,
                                                                 kernel_training='linear',
                                                                 kernel_training_param=1,
                                                                 kernel_target='linear',
                                                                 kernel_target_param=1,
                                                                 regularization=0.1)

# Print the final results
print(best_model)

Copyright & License

Copyright (c) 2018, Mindstrong Health. GNU Affero General Public License.