I have data of construction site and am wondering if i can use machine learning to reduce the cost it takes to build a building. But, as far as i know, Machine learning can only does function approximation. But, i remember that google used AI to reduce the costs of the data center. This and This are the article.
We actually wrote this for an engineering company. Basically we modeled the project business processes, then decomposed them to metrics for analysis. I can high level three cost savings areas we found.
Purchasing: use receipts, billing, and time metrics to identify outliers and reconciliation failures (receipts != payments). Outliers/missing may indicate mistakes or fraud/theft/overcharging in billing (ie:$200 hammer). Used filter functions to find.
Contract and Permit analysis: make sure all required permits and contract provisions are in place. Used text classification and boolean (exception reporting, really).
Project length prediction: if a project is shorter or longer than others in it's class, investigate. Used statistical analysis (the client already classified: public/private, plumbing/sewage, location, etc).
This was before TensorFlow, which likely could improve on our results. You will find that there are plenty of features (metrics) to model in these processes.
So hopefully this can help on model definition and consequently the objective functions.
Yes. For any machine learning problem, you are learning an objective function mapping from your input to your output. In google's case the inputs be controlling parameters and outputs be energy consumption, while in yours they are building parameters and cost.
You can do it in a standard supervised learning framework. Or you can put it in a reinforcement learning framework in which you treat the cost as reward signal, if you have enough data or have some simulation software to generate data.