This documentation provides the project workflow of the relevant bridge projects:
- Decision tree
- Actuarial life tables
The workflow for the decision tree project consists of data acquisition, data cleaning and transformation, modeling, and prediction. Figure 1 describes the five-step process of this project.
Fig. 1 Data process for the Actuarial Life Table
To acquire the NBI dataset this sections provides in detail instructions.
The NBI dataset is a temporal dataset containing data from across all states in the united states from 1992 - 2022. To build a model to predict future maintenance of the model or deterioration. We have employed several transformation techniques that take into account the format of the data.
The data generation is a separate script from the data acquisition script, and it is often customized according to the application. The generation of data also considers basic cleaning and transformation of the dataset. For each application, we have highlighted the transformation taken to reproduce the dataset.
This format takes into account the ground truth.
The actuarial life table project workflow consists of data collection, data cleaning and transformation, modeling, and prediction. Figure 2 describes the process of this project.
Fig. 2 Data process for the Actuarial Life Table
The actuarial life table project is based on the dataset acquired from the NBI dataset. However, specific transformations are described in detail within the project repository.
Similar to the Decision Tree project, the Actuarial Life Tables' data cleaning and transformation process are based on the NBI dataset. To build the bridge maintenance life tables, the project considers all data across states and years. Moreover, the dataset is transformed into time-series data to identify maintenance patterns by observing improvement in condition ratings of the bridges using the Bridge Intervention Matrix.