Purpose

We are creating a way of verifying whether a banknote is original(genuine) or forged(fake). We are using K-Means algorithm to train our model to perform this very action.

Description of the Data

We have provided a data set containing 1,372 banknotes some which are genuine and some which are false(forged).

Methods: how the data was analyzed

We used the data set provided to train our model using K-Means algorithm. This is just to ensure that our model will be able to tell the two types of banknotes apart. K-Means is a type of algorithm used in the case of unsupervised learning grouping unlabeled data set into different clusters. Each cluster will be such that a dataset will belong to only one group with which it shares similar characteristics. The K is used to define the number of clusters created as in our case we grouped them into two clusters(K=2) We imported the necessary libraries that would assist us in performing the required operations on the data e.g. putting our data into a table format, visualizing our data etc. We pulled out the 2 columns we were interested in (v1 & v2) from the data set we were provided with. Afterwards applied K-Means algorithm to the 2 columns (we had conflated into one column first) and came up with the 2 clusters we had pre-defined. We ran K-Means multiple times (9) to see if our clustering was stable. The clustering is fairly stable as we didn’t get many different results every time we rerun k-means. The results were very similar.

Summary of the result

There are 599 data being clustered in group 0 and 773 being clustered in group 1 after k-means.

Visualization of Result

Our accuracy was based on the matching values between the original class plot and the k-means plot out of the entire data set i.e. 1371 K-Means Result: The accuracy of this K-Means Model is 34.79%. Recommendations The accuracy level will affect the future results of your experiment on banknote verification. I would advise using another algorithm or find ways of optimize our accuracy rate.