Hello everyone!
I am actually looking for experts opinion regarding which machine learning algorithm is best suited for construction accident data.
Looking forward to getting your wonderful answers.
Thanks.
Could you provide some insight into the data? Depending on the data and its type different algorithms are fruitful.
Soumya Snigdha Kundu , the data comprises of date at which the accident occurred, the location, the causes or contributory factors, injury sustained, the number of construction worker involved and the types of accident. I hope this answered your question
None is the best,it depends on how much information you can use from the data.If the model has multiple assumptions,you can try Alternating direction method of multiplier (ADMM) algorithm,which is a popular machine learning algorithm suited for large scale data optimization.
Thank you for the insight! Since its table wise data I guess you can try starting off with forests or trees then go for ensembling approaches but like Yunze Yue mentions, none is the best as such in a general sense. You will also have consider how well your data is. If it is skewed in any way or greatly missing values then different algorithms might have to be accounted for.
Soumya Snigdha Kundu Missing values isn't that imoprtant,it can be seen everywhere in the project.The best algortihm based on how to bulid your math problem which can reflect your problem.For the construction of your data,you must find the way to bulid your model from the practical project,which is called Sensing Martix in Machine Learning.As long as the Sensing Martix is okay,you can use so many Compressed Sensing algorthims to solve the reconstruction problem from missing data.
Yunze Yue Thank you for the clarification! In essence what I meant was you have to look out for that and the success of the overall task is not solely dependent on the machine learning model alone.
I also meant in any way "or" and not "for". Sorry about that.
Yunze Yue , Thank you very much for your valuable and professional input on this subject matter. I am really grateful and I will try ADMM
The question is how is your data 'to learn' look alike? Is it a set of:
- photo after,
- set of photos with a defined time interval, or
- short video that shows an accident from time before to its end?
Construction accidents differ based on the source of fail:
- human,
- machine,
- static part of the construction,
- variations between them.
We are doing such a solution but it makes us train model after model and build a situation tree-based awareness decision engine.
Then there are factors that lower the quality of each assessment [like the amount of light].
well am not quite sure about your request but maybe any (HAZOP) based software might help some kind such as isograph hazop , hazop + or if possible doing some simulation using Matlab which still not that easy. or even so a support vector machine-based software. good luck in your research
- Badges
- Science topic