This is the course site for data-driven infrastructure management, with a strong emphasis on the pavement systems. This class starts core concepts of infrastructure management. It then moves to the many aspects in using data for management, including data-based maintenance policy making, performance prediction, and automated condition evaluating. We rely heavily on using R programming language for data wrangling, cleaning, visualization, and modeling. For modeling, we will both use R and Python programming languages in creating machine learning models. This class requires no background on either R and Python, although a good commanding of either will be beneficial. We will give a brief introduction to the get class started.
Prerequisite
Basic concepts on pavement design and paving materials are required, which the class Road Engineering or Pavement Engineering covers. We will use data from the Long-Term Pavement Performance database.
Syllabus
This is a xxx credit course, the syllabus comprises of three components:
1. Concepts of infrastructure Management
- Key components
- life-cycle cost analysis (LCCA)
2. Fundamental skills for applying data for infrastructure management
- Basic concepts of
Rprogramming- Workflow of using
Rfor infrastructure data management - Data wranglying using
tidyverse:tibble,dplyr - Data visualization using
ggplot2 - Processing millions of data records using
data.table
- Workflow of using
- Basic concepts of
pythonProgramming- Data processing with
numpyandpandas - Data visualization using
matplotlib
- Data processing with
- Reproducible research with literature programming: using
rmarkdown- Basic syntax of
markdown - Fundamentals of
rmarkdown - Mention
org-modeandlatex
- Basic syntax of
3. Machine-learning methods for performance prediction and decision-making
- Maintenance decision tree with
rpartandggtree - Prediction with regression methods: linear regression, mixed effects models, generalized additive models, decision trees, random forests, gradient decision tree models, multilayer preceptron, and others.
- Discrete responses prediction through classification: decision trees, logistic regression models family
