1. Abstract
The
exhaustion
of
principal
energy
resources
along
with
everyday
global
energy
management
problems
forces
measures
to
be
taken
with
the
purpose
of
finding
solutions
to
reduce
energy
consumption.
Although
the
transport
sector
is
a
major
energy
consumer,
a
perfect
tool
to
predict
the
energy
consumption
of
all
types
of
vehicles
does
not
currently
exist.
Existing
literature
describes
several
models
that
give
good
predictions
for
energy
consumption
at
a
microscopic
level,
but
none
of
them
are
versatile
enough
to
be
able
to
make
a
prediction
in
different
conditions
such
as
different
environmental
and
car
specifications.
In
this
study,
a
new
model
is
introduced
using
a
random
forest
algorithm
with
the
capability
to
predict
energy
consumption
for
light
and
heavy-‐duty
fuel
vehicles.
Included
is
a
description
of
a
series
of
tests
performed
on
the
model
to
analyse
the
robustness
of
random
forest,
such
as
cross-‐validation,
on
energy
consumption
prediction.
It
is
expected
that
with
this
type
of
machine
learning
algorithm
the
energy
consumption
prediction
becomes
more
accurate
and
consequently
one
can
produce
a
model
capable
of
performing
predictions
for
other
types
of
vehicles
(hybrid,
electrical)
in
any
location
on
the
globe.