wall-ml-prediction 1.0.1

This library presents available functions to use Artificial Neural Netwroks (ANN) and Random Forest (RF) trained models to make predictions of the maximum roof acceleration (PFA_max) and the maximum roof drift ratio (rDR_max) in reinforced concrete wall buildings.

ML prediction - RC Wall buildings

Note: This is a README file adapted from the available GitHub repository at https://github.com/carlosantr/Wall_ML_Prediction

This repository presents the trained ML models of the paper "Machine Learning based evaluation of roof drifts and accelerations of RC wall buildings during ground motions". Artificial Neural Netwroks (ANN) and Random Forest (RF) are available to make predictions of the maximum roof acceleration (PFA_max) and the maximum roof drift ratio (rDR_max) in reinforced concrete wall buildings. This repository presents the trained models, prediction functions (please see the function documentation in the .py file) and examples for multiple and individual prediction modes. You can use the prediction functions in your own project with two different ways:

1. Install the open-source library distributed through Python Package index (PyPi) using the command "pip install wall_ml_prediction" in your local command prompt (Anaconda prompt is recommended). Once the installation is done, you can call the prediction functions in your projects with the command "from wall_ml_prediction.predict import prediction_ML_walls_individual" or ""from wall_ml_prediction.predict import prediction_ML_walls_multiple". Then, use freely the functions to make predictions of PFA_max or rDR_max using the trained ML models (i.e. RF and ANN).

2. Download this repository and locate the files (in some cases you will need to unzip the downloaded file from the repository) in the same path of your project. Then, call the functions from the predict.py file as follows "from predict import prediction_ML_walls_individual" or ".predict import prediction_ML_walls_multiple". Then, use freely the functions to make predictions of PFA_max or rDR_max using the trained ML models (i.e. RF and ANN).

For further information about the input data required by the functions, please see the documentation in the predict.py file in this repository. The abovementioned documentation is presented below:

Individual prediction mode:

• WI: Wall Index
• T1: Fundamental period
• H_Tcr: Stiffness Index
• Ar: Mean wall aspect ratio
• ALR_G: Average axial load ratio for gravity loads
• AI: Arias Intensity
• Sa: Spectral acceleration
• Sv: Spectral velocity
• var_predict: Is a optional variable, and must be a list with the output values to predict (i.e. ["PFA_max", "rDR_max"], ["PFA_max"] or ["rDR_max"])
• model_predict: Is an optional variable, and must be a list with the Machine Learning models to do the predictions (i.e. ["RF", "ANN"], ["RF"] or ["ANN"])

Multiple prediction mode:

• X must be a dataframe with the next columns (each row is a prediction): WI: Wall Index T1: Fundamental period H_Tcr: Stiffness Index Ar: Mean wall aspect ratio ALR_G: Average axial load ratio for gravity loads AI: Arias Intensity Sa: Spectral acceleration Sv: Spectral velocity
• The columns must have exactly the same name (i.e., X.columns = ['WI', 'T1', 'H_Tcr', 'Ar', 'ALR_G', 'AI', 'Sa', 'Sv'])
• var_predict: Is a optional variable, and must be a list with the output values to predict (e.g. ["PFA_max", "rDR_max"], ["PFA_max"] or ["rDR_max"])

IMPORTANT NOTE:

• The functions are expected to be used with input data in the range presented in Table 1 of the paper. If you use out-of-range values, the predictions could be worst. Be careful about the data you input to the prediction functions.
• For the second case (downloading directly from the repository), you have to accomplish with the dependencies requirements in your python environment (Scikit-Learn: write in console: pip install --user scikit-learn==1.2.1, Joblib: write in console: pip install --user joblib==1.4.2, Keras: wite in console: pip install --user keras==3.5.0, and Tensorflow: write in console: pip install --user tensorflow==2.17.0.