Author has tried to understand Sentiment Analysis approach used by kount.com and shared his view on the future of Machine Lernning .
In this era of the internet, can we think of life without digital payment? Digital payments are key for eCommerce business. It’s not only overhead for the company to deal with Fraud transaction but also it hampers its brand, fine, customer and extra effort to clear the problem.
Nowadays most of the e-Commerce firm work with the third party which validates the transaction and approves on the fly in a fraction of seconds. Such providers use machine learning to offer the service, not only that some of such firms (not talking about kount.com) they also offer to pay back to the company if any fraud occurs.
The (business) domain under discussion: E-commerce, Digital Payment
The problem that sentiment analysis addresses: Control Fraud Transaction, identify such transaction and report them.
Approach used by kount.com
In this case, I am talking about the approach used by kount.com, they are using Supervised and Unsupervised approach to solve this problem.
Supervised: Supervised approach gives the flexibility of making the decision based on the trusted models /decision tree. And report any anomaly to the customers.
Unsupervised: This approach is used to block or monitor the known threat, like some CARD or Some location or some metadata that is very common in Fraud transaction. This helps to avoid or reduce such transaction without going through a big cost cycle for their customers.
Describe if possible a different domain where the same approach might also be effective.: Predicting election fraud, or dropping such votes. Not sure how it can be practically implemented but if data points are accessible then some model can b built to avoid such fraud in the future.
want to read more : https://www.kount.com/how-fraud-prevention-works/machine-learning