Build a Neural Network model to predict the type of incident. Discuss the accuracy and some of the challenges that you had.

Problem:

The city of Austin has heard many complaints from cyclists that the city isn’t doing enough to protect them from motor vehicles. To confirm these complaints, the city has compiled data of cyclists related incidents and want you to review these findings to see if the data confirms this. Pick 6 independent variables and explain why you choose those.

Part 1:

Build a Logistic Regression model to predict, Severity ( Possibly Injury, Incapacitating Injury, Killed vs. a non severe incident). Talk about which features were important from the model and if the model is accurate.

Part 2:

Build a Random Forest model to predict the type of incident. Compare which metrics are important in this model vs significant in the Logistic Model and how the model compares. Run with minimum trees of 5000

Part 3:

Build a Neural Network model to predict the type of incident. Discuss the accuracy and some of the challenges that you had.

Part 4:

Using multiple benchmarking metrics (as covered in class) to compare and contrast both models. Summarize your findings and suggest a final model for the city to use. Focus on business interpretation, but your conclusion should reveal your reasoning on why/why not this model could be used based on benchmarking metrics.