[SOLVED] Hive ECE 720 PROJECT

ECE 720 PROJECT
Winter 2017

Deadline: The report to be submitted on the last day of classes of the Winter semester.

Consider one-dimensional time series coming from the UCI machine learning repository
http://archive.ics.uci.edu/ml/
or KEEL
http://sci2s.ugr.es/keel/datasets.php
The time series under discussion belong to one of several classes. The data set is split into training and testing set (60-40). The training data are used to develop a classifier.
The series is quantized in amplitude space by using
(a) equal length intervals spread between min and max, and
(b) equal probability intervals
60-40

a
b
As a result, the series is represented as a sequence of visible symbols (states), say ABCCDFA
ABCCDFA
Design HMMs for this classification problem. Assume that each sequence starts with a null symbol and ends with an end null symbol. Present a thorough analysis of the obtained HMMs and analyze its performance. In particular, show-transition matrices
HMMHMM

-analyze an impact of the number of hidden states and the number of symbols on the performance of the classifier
-classification performance obtained on the training and testing data

In your report include a code along with its detailed commenting. Describe the data set used.

/docProps/thumbnail.wmf

ECE
720
PROJECT
Winter
2017
Deadline:
The
report
to
be
submitted
on
the
last
day
of
classes
of
the
Winter
semester.
Consider
one-dimensional
time
series
coming
from
the
UCI
machine
learning
repository
http://archive.ics.uci.edu/ml/
or
KEEL
http://sci2s.ugr.es/keel/datasets.php
The
time
series
under
discussion
belong
to
one
of
several
classes.
The
data
set
is
split
into
training
and
testing
set
(60-40).
The
training
data
are
used
to
develop
a
classifier.
The
series
is
quantized
in
amplitude
space
by
using
(a)
equal
length
intervals
spread
between
min
and
max,
and
(b)
equal
probability
intervals

60-40

a

b

As
a
result,
the
series
is
represented
as
a
sequence
of
visible
symbols
(states),
say
ABCCDFA

ABCCDFA

Design
HMMs
for
this
classification
problem.
Assume
that
each
sequence
starts
with
a
null
symbol
and
ends
with
an
end
null
symbol.
Present
a
thorough
analysis
of
the
obtained
HMMs
and
analyze
its
performance.
In
particular,
show-transition
matrices

HMM

HMM

-analyze
an
impact
of
the
number
of
hidden
states
and
the
number
of
symbols
on
the
performance
of
the
classifier

-classification
performance
obtained
on
the
training
and
testing
data

In
your
report
include
a
code
along
with
its
detailed
commenting.
Describe
the
data
set
used.