The following explains how to build in Python a decision tree regression model with the FARS-2016-PROFILES dataset.

Here, the purpose is to get some prediction for the 4 following crash profiles that do not exist in the « FARS-2016-PROFILES » dataset :
According to 2016 data, we want an estimation of
1) the number of deaths in a road crash located in a completely dark (2) rural (1) road of Texas (48) occurring a rainy (2) friday (6) involving 2 vehicles 4 people and 1 drunk driver.
2) the number of deaths in a road crash fitting the same previous profile but without any drunk drivers
3) the number of deaths in a road crash located in a completely dark (2) rural (1) road of California (6) occurring a rainy (2) friday (6) involving 2 vehicles 5 people and 1 drunk driver.
4) the number of deaths in a road crash fitting the same previous profile but without any drunk drivers

We start by importing the needed libraries and loading the dataset :

#importing libraries
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd

#loading the dataset
dataset = pd.read_csv('FARS-2016-PROFILES.csv')
X = dataset.iloc[:,0:8].values
y = dataset.iloc[:,len(dataset.iloc[0])-1].values


The dataset is an array whose 30 first rows (over 20083) are as follows :

FARS 2016 « Profiles » Dataset in Python : Number of deaths in road crashes according to particular conditions

With the following code we build the Decision Tree Regression Model with the dataset :

#fitting the Decision Tree Regression model to the dataset
from sklearn.tree import DecisionTreeRegressor
regressor = DecisionTreeRegressor()
regressor.fit(X, y)


The following code answers to the first two questions in Texas state :

#prediction of the number of deaths in a car accident in Texas (48), a friday (6), during a night without lights (2), when raining (2), on a rural road (1) involving 2 vehicles 4 people and 1 drunk driver.
y_pred_TX_Drunk = regressor.predict([[48,6,2,2,1,2,4,1]])

#prediction with the same parameters but no drunk drivers (all drivers were sober).
y_pred_TX_Sober = regressor.predict([[48,6,2,2,1,2,4,0]])


After running this script, we get y_pred_TX_Drunk=3 and y_pred_TX_Sober=1. Therefore, we get the prediction that
1) IF a road crash located in a completely dark (2) rural (1) road of Texas (48) occurres during a rainy (2) friday (6) involving 2 vehicles 4 people and 1 drunk driver THEN according to 2016 data, the estimated number of deaths is 3.
2) IF a road crash occures in the same conditions without any drunk drivers THEN according to 2016 data, the estimated number of deaths is 1.

The following code answer to the next two questions in California state :

#prediction of the number of deaths in a car accident in California (6), a friday (6), during a night without lights (2), when raining (2), on a rural road (1) involving 2 vehicles 5 people and 1 drunk driver.
y_pred_CA_Drunk = regressor.predict([[6,6,2,2,1,2,5,1]])

#prediction with the same parameters but no drunk drivers (all drivers were sober).
y_pred_CA_Sober = regressor.predict([[6,6,2,2,1,2,5,0]])


After running this script, y_pred_CA_Drunk=2 and y_pred_CA_Sober=1. Therefore, we get the prediction that
1) IF a road crash located in a completely dark (2) rural (1) road of California (6) occurres during a rainy (2) friday (6) involving 2 vehicles 5 people and 1 drunk driver THEN according to 2016 data, the estimated number of deaths is 2.
2) IF a road crash occures in the same conditions this time without any drunk drivers THEN according to 2016 data, the estimated number of deaths is 1.

We can visualize a part of the decision tree (here, the first 5 levels) with the following code :

#visualizing decision tree
from sklearn.externals.six import StringIO
from IPython.display import Image
from sklearn.tree import export_graphviz
import pydotplus

export_graphviz(regressor, out_file=dot_data, max_depth=5,
filled=True, rounded=True,
special_characters=True)
graph = pydotplus.graph_from_dot_data(dot_data.getvalue())
Image(graph.create_png())


This script creates the following PNG image :

The first 5 levels of the decision tree regression model