Introduction to Feature Engineer
Practical cost forecast is currently becoming a severe problem due to the Covid-19-related
disruptions such as supply chain management, local labor shortage, and escalated inflation.
Luckily, emerging data science integrated with construction data has the potential to become
the evidence-based solution for cost estimation because Machine Learning (ML) can
significantly improve the accuracy of the predictions based on the identified statistical patterns
(Burns, 2021). Feature engineering is the fundamental technique of ML to look up, select, and
analyze the generated data. The measurable feature can be inputted into the algorithm,
ensuring the designed artificial intelligence can help the predictive model complete the data
transformation from the observed trend or variables (Patel, 2021).
The following essay will discuss how the featured engineering can help our client, the Texas
Department of Transportation(TxDOT), to effectively evaluate the bidding documents and how
to implement ML technology to facilitate the company's decision-making strategy.
Why does TxDOT Need Feature Engineering
Our client, TxDOT's asset inventory, comprises aviation, public transportation, and the state
highway system (Texas Department of Transportation, 2022) . Our research study will only
focus on the operation & maintenance(O&M) of the state highway system due to the limited
space of the essay.
According to the Texas Transportation Plan 2040, the TxDOT has approximately 314,000 miles
of public roadways, and the O&M budget for highway pavement in Texas is estimated at $103.7
billion until 2040, which is $4 billion annually in 2014 dollars value. Since TxDOT receives nearly
50 percent of its funding from performing regular O&M for the state highway by the federal
Highway Trust Fund (HTF), a highly accurate budget plan is critical to getting approved by the
funding authority (Kevin McPherson, 2018). There are three central components for sustainable
asset management plans: estimated road usage, deteriorating model of the current asset, and
fluctuating construction costs. The following paragraphs will explain the importance of these
three dependent variables to understand why feature engineering should efficiently manage
these three variables.
, Estimated road usage: The usage of passengers is one of the biggest causes of wear and tear
problems in road construction. Furthermore, the rapid population growth can drive up the
congestion time, potentially accelerating the roadway system's deterioration rate. The usage
rate can be measured by the daily vehicle traveling distance in miles, which increased by
around 16 percent between 2010 and 2016 and is highly associated with significant population
growth and economic activity in Texas (Kevin McPherson, 2018). The current model predicts
that the Texas population will increase to 45 million, and vehicle travel will be 800 million miles
per day in 2040. Feature engineering in reinforced learning should closely monitor the newly
published population size and update modeling to increase the accuracy of the estimated
usage. (Costello, 2016)
Deterioration Rate Model: The pavement maintenance management system relies heavily on
the deterioration rate prediction model, which can help the owner set up workable financial
budgeting based on the life cycle economic analysis. The mechanistic-empirical model is the
most widely adopted for the asset management agency since it can provide an insightful
prediction for the future road condition based on the collected physical property of the road
section and climatic condition, including functional class, the number of remedial actions,
freeze index, etc. (K.P George, N/A) . Feature engineering can help the algorithm clean up the
corrupted data by auto-filling with the most similar numbers from the established classification
modeling. Eventually, the enhanced artificial intelligence can help the asset management team
decide to screen out unnecessary features relevant to the future data collection process.
Construction Cost: Unit price bidding is a typical bidding contract for horizontal construction
since the quantity of the performed work can be verified by the owner. It will provide a cost
breakdown of the items, including material, labor and equipment, and so forth. This type of
contract usually will have the specified clause for the contingencies to bond with some
commodity index for construction material such as timber price, oil price, etc. These indexes
usually have high volatility due to huge price fluctuation, so general contractors usually add up
some additional premium to cover the price risk in the fixed-price contract. Additionally, the
current inflation price of construction materials rose by 31.3 percent in 2022 from 2020, the
highest rise in the last 40 years (Mosier, 2022). The featured engineering can save enormous
Practical cost forecast is currently becoming a severe problem due to the Covid-19-related
disruptions such as supply chain management, local labor shortage, and escalated inflation.
Luckily, emerging data science integrated with construction data has the potential to become
the evidence-based solution for cost estimation because Machine Learning (ML) can
significantly improve the accuracy of the predictions based on the identified statistical patterns
(Burns, 2021). Feature engineering is the fundamental technique of ML to look up, select, and
analyze the generated data. The measurable feature can be inputted into the algorithm,
ensuring the designed artificial intelligence can help the predictive model complete the data
transformation from the observed trend or variables (Patel, 2021).
The following essay will discuss how the featured engineering can help our client, the Texas
Department of Transportation(TxDOT), to effectively evaluate the bidding documents and how
to implement ML technology to facilitate the company's decision-making strategy.
Why does TxDOT Need Feature Engineering
Our client, TxDOT's asset inventory, comprises aviation, public transportation, and the state
highway system (Texas Department of Transportation, 2022) . Our research study will only
focus on the operation & maintenance(O&M) of the state highway system due to the limited
space of the essay.
According to the Texas Transportation Plan 2040, the TxDOT has approximately 314,000 miles
of public roadways, and the O&M budget for highway pavement in Texas is estimated at $103.7
billion until 2040, which is $4 billion annually in 2014 dollars value. Since TxDOT receives nearly
50 percent of its funding from performing regular O&M for the state highway by the federal
Highway Trust Fund (HTF), a highly accurate budget plan is critical to getting approved by the
funding authority (Kevin McPherson, 2018). There are three central components for sustainable
asset management plans: estimated road usage, deteriorating model of the current asset, and
fluctuating construction costs. The following paragraphs will explain the importance of these
three dependent variables to understand why feature engineering should efficiently manage
these three variables.
, Estimated road usage: The usage of passengers is one of the biggest causes of wear and tear
problems in road construction. Furthermore, the rapid population growth can drive up the
congestion time, potentially accelerating the roadway system's deterioration rate. The usage
rate can be measured by the daily vehicle traveling distance in miles, which increased by
around 16 percent between 2010 and 2016 and is highly associated with significant population
growth and economic activity in Texas (Kevin McPherson, 2018). The current model predicts
that the Texas population will increase to 45 million, and vehicle travel will be 800 million miles
per day in 2040. Feature engineering in reinforced learning should closely monitor the newly
published population size and update modeling to increase the accuracy of the estimated
usage. (Costello, 2016)
Deterioration Rate Model: The pavement maintenance management system relies heavily on
the deterioration rate prediction model, which can help the owner set up workable financial
budgeting based on the life cycle economic analysis. The mechanistic-empirical model is the
most widely adopted for the asset management agency since it can provide an insightful
prediction for the future road condition based on the collected physical property of the road
section and climatic condition, including functional class, the number of remedial actions,
freeze index, etc. (K.P George, N/A) . Feature engineering can help the algorithm clean up the
corrupted data by auto-filling with the most similar numbers from the established classification
modeling. Eventually, the enhanced artificial intelligence can help the asset management team
decide to screen out unnecessary features relevant to the future data collection process.
Construction Cost: Unit price bidding is a typical bidding contract for horizontal construction
since the quantity of the performed work can be verified by the owner. It will provide a cost
breakdown of the items, including material, labor and equipment, and so forth. This type of
contract usually will have the specified clause for the contingencies to bond with some
commodity index for construction material such as timber price, oil price, etc. These indexes
usually have high volatility due to huge price fluctuation, so general contractors usually add up
some additional premium to cover the price risk in the fixed-price contract. Additionally, the
current inflation price of construction materials rose by 31.3 percent in 2022 from 2020, the
highest rise in the last 40 years (Mosier, 2022). The featured engineering can save enormous
