fmegahed/shoulder_fatigue_modeling

Shoulder Fatigue Modeling

Introduction

This is the code used to perform the analyses discussed in [paper name here]. The primary goals of this research paper are:

• Can wearable sensor features be used to classify localized fatigue development for the upper extremity?
• How many wearable sensors, and at which locations, are needed to obtain satisfactory predictive performance of fatigue?
• What are the important predictive features to obtain satisfactory predictions?
• Can models be effectively generalized across subjects, or is there a need for individualized subject models?

The analyses performed to answer the questions posed are all in the analysis.ipynb notebook.

Setup

IMPORTANT: A Python virtual environment is highly recommended. You can learn how to create a virtual environment here. Different operating systems have different methods necessary to create a virtual environment.

This project uses Python 3.9.12 (download here). To run the notebook files in this project, you will need to install pycaret and ipympl. To do this, run the commands below in your terminal:

$ pip install pycaret[full]
$ pip install ipympl

Once it is done downloading, you should be able to run all of the Jupyter Notebook files without issue.

Project Directory Structure:

The data that is used to generate all of the results is stored in the data/ directory. The file that is used is initial_features_mod.csv. This file is a modified version of the initial_features.csv file, with additional rows to make model generation simpler. The resulting data generated from the code is stored under the outputs/ directory and its subdirectories.

The directory tree should look as follows:

📁shoulder_fatigue_modeling/
├── 📁data/
│   ├── 📄initial_features_mod.csv
│   └── 📄initial_features.csv
├── 📁m_creation/
│   ├── 📁analysis/
│   │   ├── 📄analysis-feature_importance.ipynb
│   │   ├── 📄analysis-subject_influence.ipynb
│   │   └── 📄analysis.ipynb
│   ├── 📁outputs/
│   │   ├── 📁general/
│   │   ├── 📁graphs/
│   │   ├── 📁individual/
│   │   └── 📁summary/
│   ├── 📄save_models_general.ipynb
│   ├── 📄save_models_individual.ipynb
│   ├── 📄save_sensors_general.ipynb
│   └── 📄save_sensors_individual.ipynb
├── 📁m_creation-classification/
│   ├── 📁outputs/
│   ├── 📄analysis.ipynb
│   ├── 📄create_data.ipynb
│   └── 📄save_general_classification.ipynb
├── 📁m_creation-general_limited/
│   ├── 📁outputs/
│   ├── 📄analysis.ipynb
│   └── 📄save_limited_model_classification.ipynb
├── 📜.gitignore
└── 📙README.md

If any of these files are missing, please download them from the GitHub repository. If any of the directories are missing, please add them following the guide shown above. The output directories will

Notebooks

The information below is a brief summary of what each notebook does. Each notebook will contain inside it additional details about its functionality.

save_models_general.ipynb

The save_models_general.ipynb notebook generates the general training and testing models, and saves them in CSV files in the directory outputs/general/. The files generated are called train_general.csv and test_general.csv, which store the training models and testing models, respectively.

The generation of the models will also provide feature importance data for all of the training models. The feature importance data will also be stored in the outputs/general/ directory. The CSV file the feature importance data is stored in is feature_importance_general.ipynb.

save_models_individual.ipynb

The save_models_individual.ipynb notebook generates the individual training and testing models, and saves them in CSV files in the directory outputs/individual/. The files generated are called train_individual.csv and test_individual.csv, which store the training models and testing models, respectively.

The generation of the models will also provide feature importance data for all of the training models. The feature importance data will also be stored in the outputs/individual/ directory. The CSV file the feature importance data is stored in is feature_importance_individual.ipynb.

save_sensors_general.ipynb

The save_sensors_general.ipynb notebook generates the general training and testing models, and saves them in CSV files in the directory outputs/general/. This notebook generates models by dividing features by sensor and generating a separate model for each sensor. The files generated are called sensor_train_general.csv and sensor_test_general.csv, which store the training models and testing models, respectively.

The generation of the models will also provide feature importance data for all of the training models. The feature importance data will also be stored in the outputs/general/ directory. The CSV file the feature importance data is stored in is sensor_feature_importance_general.ipynb.

save_sensors_individual.ipynb

The save_sensors_individual.ipynb notebook generates the individual training and testing models, and saves them in CSV files in the directory outputs/individual/. This notebook generates models by dividing features by sensor and generating a separate model for each sensor. The files generated are called sensor_train_individual.csv and sensor_test_individual.csv, which store the training models and testing models, respectively.

The generation of the models will also provide feature importance data for all of the training models. The feature importance data will also be stored in the outputs/individual/ directory. The CSV file the feature importance data is stored in is sensor_feature_importance_individual.ipynb.

The analysis.ipynb file contains many sections that analyze different aspects of the generated models.

individual_vs_general.ipynb

The individual_vs_general.ipynb notebook provides a simple comparison between the individual and general models. For each subject, the average of the general models is compared to its individual counterpart.

analysis.ipynb

The analysis.ipynb notebook contains several analyses performed on the model results to determine how models, features, subjects and more perform in predicting subject fatigue. The analyses conducted are listed below.

• Subject Influence: The goal is to determine whether (in the General models) certain subjects result in models that are more or less effective at predicting fatigue by determining how frequently they appear in the training sets of models that perform well or poorly
• Most Important Features: The goal is to find which features are the most influential in high-performing models.
• Plotting Feature Importance: The goal is to see how all of the different features compare to each other on a graph.
• Graphing MAE: The MAE values of General models and Individual models will all be graphed on Box, Scatter and Violin plots to determine which ones perform the best.
• Model Summary: The General and Individual model tables are summarized to determine which models perform the best for each subjects, as well as what their min, median and max MAE values look like.

File Execution Order:

There is a recommended order in which to run the files to make sure that everything works as intended. This is listed below in steps:

Repurpose descriptions from here to Files and Notebooks?

1. Run the save_models_general.ipynb and save_models_individual.ipynb files.

2. Run the save_sensors_general.ipynb and save_sensors_individual.ipynb files.

3. Run the analysis.ipynb file.

If the general/, individual/ and summary/ folders already contain the necessary CSV files, it is not necessary to run the save_models_general.ipynb, save_models_individual.ipynb, save_sensors_general.ipynb and save_sensors_individual.ipynb files unless you want to regenerate the models.

These files are likely already provided because it can between several hours and a few days to run the model generation notebooks depending on your machine.