Driver behavior Cluster

Introduction

In this project, we tried to create a statistical model to cluster driver behavior based on CAN Bus sensors data.
We will use Hierarchical clustering to identify and group the drivers based on their behavior and driving style. This identification of drivers can be used for improvements.

Data preparation

overview.csv
the datasets, contains 42 parameters (columns) and 60 variables (observations),

Data cleaning

Before moving to data analysis, we need to clean our dataset:
converting types, replacing missing values with zeros.

By plotting the coorelation matrix, we can consider the variables with the lowest coorelation coefficient, which are the ones explaining the variability. Also, this step will allow us to reduce the number of parameters to consider in our analysis.

Features

id: identifier of the vehicle.
odo: The odometer reading from the vehicle in km.
dist: Driven distance during the time period.
fuelc: Total fuel consumption during report period while driving, idling and using a power take-off (litres).
idle: Engine running time in idle mode expressed as HH:MM:SS
pause: Engine running time with pause expressed as HH:MM:SS
fuelr: How many litres of fuel the vehicle or driver has consumed per 100 km.

The learning model

Using the hierarchical clustering, to identify clusters within the population.