Scribd
Upload
18 views

Data Science Project

The document discusses using a K-Means clustering algorithm and data science to detect fake banknotes by measuring the variance and skewness of wavelet transformed images and representing the correlation between these metrics in a graph. The algorithm identifies where data points cluster to predict whether a banknote is legitimate or fake based on its proximity to the identified cluster points.

Uploaded by

Juan Ignacio Viviani
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
18 views

Data Science Project

The document discusses using a K-Means clustering algorithm and data science to detect fake banknotes by measuring the variance and skewness of wavelet transformed images and representing the correlation between these metrics in a graph. The algorithm identifies where data points cluster to predict whether a banknote is legitimate or fake based on its proximity to the identified cluster points.

Uploaded by

Juan Ignacio Viviani
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 3

Data Science Project

Report

Our purpose in this project is trying to detect fake


banknotes through a data science algorithm.
The purpose is predict if a banknote is legit or not giving a
result. For this we use an algorith known as “K-Means”
which gives us a result based on the proximity on K
points, for knowing the result, if it’s genuine or a fake
banknote, we are going to explain it better next in this
report.

We start with some data information, we know two


variables or things that we can measure for our purpose,
“V1” and “V2”.
“V1” means variance of Wavelet Transformed image,
which is a mathematic technique to analyze images when
we measure the variance.
Now, “V2” means skewness of Wavelet Transformed
image, who is used to measure the asymmetry of
differents images (in this case, we are going to analyze
banknotes).

Now, depending on V1 and V2 we can classify banknotes


in two differents Classes: We are going to consider if it’s
in Class 1 the banknote is genuine, and if it is fake, it’s on
Class 2.

We measure V1 and V2 in a normalized o standard form,


because both of them have different scales, so we don’t
work properly with the data if we are going to using
different scales to measure and analyze. So first of all, we
normalize the two variables finding the maximum and the
minimum of V1 and we do it the same with V2, next we
apply a mathematic formula to standarize both data.
Now we have properly scaled data, we apply the K-Means
algorith, as we say it early, the K-Means it’s a technique
who finds where are the data clustering on. Now, this
areas where the data is clustering, are marked by a point
on the graph.

Explaining the graph:

This graph, even if looks very terrifying at first view, it is


very simple.

On the first axis, X, we have the data of “V1”


and the second axis, Y, we have the data of “V2”.
In this graph all the blue points means the correlation
between each other (V1,V2) we can call that a vector.
As we can see, the blue points make a figure, our work
now is trying to find the K-Means clusters.
If you take attention, you can probably guess that red
points are K-Means cluster, and you are right.
The red points is where our data science algorithm finds
clusters in all this vectors or (V1, V2) points, which for us it
get it so much difficult to find this.

Now, in summarize, we can say:

- Our purpose is trying to identify fake banknotes, for this


we use Data Science algorithms, specifically K-Means
Clustering.
- We measure the variance of Wavelets transformed
called V1 and skewness of Wavelets transformed called
V2, which are mathematical techniques for analyze
images of banknotes in this case.
- We represent the correlation between V1 and V2 in a
graph and we apply the K-Means algorithm which help us
to find where the points are clustering on.

Now, maybe you’re probaly thinking, ok, that’s great but


we can predict now if a banknote is legit or not?
Well, the answer is yes, so knowing the K-Means clusters
if a vector is nearer at the bottom-left point, we can say the
banknote is probably genuine and if it’s nearer for the top-
right point, it’s a fake banknote.

Recomendations:

With this technique, you can make a computer analyze the


banknotes and check if the vector is nearer of which point,
the result gives you an answer “Genuine” or “Fake”
banknote. It’s important to consider that this algorithm is
not 100% effective but is highly certain. It’s important to
keep developing some other techniques and data science
algorithms to detect and predict fake banknotes. We know
that bad guys are improving their technologies to make
more “genuine look” of banknotes, it’s important to be a
step ahead of them.

You might also like