I am really fascinated by the capabilities of deep learning and future evolutions of it (http://deeplearning.net/). Google, Facebook and Neflix understand its capabilities and investing a significant amount of money in future developments. As a scheduler I see deep learning as a way to create hypothesis about schedule completion dates besides using CPM logic and resource management. I believe an evolving algorithm considering more project factors and information would likely create another tool to predict a project finish date.
One of the things I believe could help the scheduling community now and also in the future is an online schedule library of old schedules. This library would help new schedulers learn faster, help current schedulers check estimates and also provide more data for AI capabilities in the future.
I would like to propose a payment incentive for people or organizations who upload schedules. While this could take on many forms, some ideas include getting paid per activity downloaded if you just want to see the air handler portion of the schedule instead of the foundation portion. Another case would be looking at a schedule that went to claims. Maybe other marketplace driven incentives as well that could evolve over time. Such as a more well known credible scheduler/organization uploading their work and charging more than a less well know scheduler/organization. Another option is a profit sharing membership method. This area has a lot of possibilities and would enjoy feedback about this. Basically instead of an old schedule sitting on a hard drive somewhere, it could be useful to someone and possibly making money.
I believe one of the biggest constraints is schedules are owned by companies and organizations. Not the scheduler who created the schedule. While this is a problem, I believe it can be minimized by stripping all company related information or just showing certain segments of a schedule like a segment about air handlers only.
Another constraint is every project is unique and different. Comparing one schedule to another is like comparing apples to oranges to a certain degree. Some schedulers use Microsoft Project while others use Primavera P6. However I believe as more schedules are uploaded, with time more information will be able to create more effective algorithms and hypothesis for project completion dates by industry, project size, etc.
I am really interested in hearing about scheduling communities response to this. Is this a good idea? Do you have a better idea? What are some of limitations not discussed in this message? I would appreciate any feedback.
Thank You,
Aaron Melton
Replies