I'm doing research on Road Accident Severity prediction using a machine learning - Artificial Neural Network. Dataset size will be less than 5000 traffic records. Thanks
The answer depends on what features/attributes you will extract. In order to get a better answer post at least number, type and corelation of attributes.
As for starting point you can try some general ensemble algorithm like random forest but depending on your dataset you can get better results with other.
Paul Stefan Popescu My dataset attributes are Place, Number of vehicle, Accident Date, Time (24hr), Time ,Type of Vehicle, Casualty Class, Casualty Severity, Sex of Casualty, Cause of accident
Sample Data=> South Okkalapa Township, 1,1-Jul-18 2:30 Night, Private car, Pedestrian, Slight ,Female, Reckless of Driving,
Having known the features/attributes is a good start. I would perform some Exploratory Data Analysis and see the results. Based on that I would choose the type of model you want to fit in. In the case of the predictive algorithm, model selection itself is the entire task.
So, it would be difficult to answer your question IMO. Based on your EDA results and hit-and trial models, checking the accuracy, mean square errors, will probably tell you the best model that fits your data. Asking which model would best fit would entirely give you guess the answer based on personal experience which will again have a high bias.
- Badges
- Science topic