Figure out the fraud ids in a bank credit card dataset. Self-Organizing Maps have been used for clustering.
SOM is an unsupervised deep learning model, mostly used for feature detection or dimensionality reduction. They output a 2D map for any number of indicators. In present times where staggering amount of data is collected with many indicators, SOMs are one of the most robust clustering techniques we have.
Weights are just co-ordinates of the output nodes in the input space. There is no activation function and summation.
Each input data point has a BMU. The BMU along with all of the neurons in its radius get updated.
Finally a 2D map is the output. A colour coding is used to identify any specefic group of data points.
I have written a detailed article on SOM here
- Clone this repository to your computer.
- Get into the folder using cd Self_Organizing_Map.
- pip install pandas
- Use minisom.py as the module for this project
Run each cell in the SOM.ipynb file.
A list of fraudulent ids who applied for bank credit cards