Time is one of the most imported input for a machine learning design to predict accidents since there is a strong correlation between time of day or even day of week to car accidents. You will use R.N.N (Recurrent Neural Network) as your only constraint, you can choose all other parameters for your liking and to your best understanding.
For more information, please refer to the following two files first: Documentation.md and Mathmatitical fundation of RNN.pdf