Fraud Detection Through ML by Ravelin.

What is Artificial Intelligence?

Artificial Intelligence is defined as the task, that is done by machine in such a way that when someone looks at the results, they feel like this is done my a human being. Thus, It clearly means that Artificial Intelligence is the state of the machine to perform task as similar as a human being would do. The ability provided to machines to think like humans is Artificial Intelligence.

What is Machine Learning ?

Machine Learning on the other hand, can be considered a subset of Artificial Intelligence. Machine learning is the concept that helps machine to learn from the data. This is done by finding a pattern in the given dataset. Thus, the performance completely depends on Data. We can say that we provide machine large amount of data and expect it to predict what a human mind does in a specific scenario (related to the provided data only).

There are two types of learning, Supervised learning and Unsupervised learning. Supervised learning involves learning through data that contains predictors as well as targets while Unsupervised learning includes only predictors while the machine has to decide the target.

Similarly, Just Like Machine Learning is the subset of Artificial Intelligence, Deep Learning is the subset of machine learning that uses neural networks to predict the outcome. Deep learning also work on large amount of data just like machine learning algorithms do. this is the reason why deep learning is the subset of machine learning.

The fields in which machine learning is widely used are-

1. Traffic Systems(google maps)
2. Social Media(Facebook/instagram)
3. Transportation and commuting(Uber/ola)
4. Product recommendations(Amazon/flipkart)
5. Virtual Personal assistant(Siri/Cortana/Google Assistant)
6. Self Driving cars (Tesla)
7. Dynamic Pricing(ola/uber)
8. Translation (google translate)
9. Online video streaming(Netflix/Prime Videos)
10. Fraud Detection(Raveline)

and so on……

Machine Learning in fraud detection

Experts predict online credit card fraud to soar to a whopping $32 billion in 2020. That’s more than the profit made by Coca Cola and JP Morgan Chase combined. That’s something to worry about. Fraud Detection is one of the most necessary Applications of Machine Learning. The number of transactions has increased due to varieties of payment channels — credit/debit cards, smartphones, numerous wallets, UPI and much more. At the same time, the amount of criminals have become adept at finding loopholes.

Whenever a customer carries out a transaction — the Machine Learning model thoroughly x-rays their profile searching for suspicious patterns. In Machine Learning, problems like fraud detection are usually framed as classification problems(Supervised learning).

Classification problem means classifying the record in a particular class. for example, determining whether the mail is an spam or not.

How does Machine Learning works for fraud detection ?

There are basically 4 steps to create a machine learning model.

1. Input- Every machine learning model needs data to analyze, which in future is responsible for the predicting whether the entry is a fraud or not. In the case of machine learning, the data should be diverse and sufficient.
2. Extract Feature- Features explains customer behaviour and fraud’s behavior to the machine. According to Ravelin, They group features into 5 main categories- Identity, order, Payment methods, Locations and Network.
3. Train Algorithm- Algorithms are a set of rules to be followed for solving a complex problem like a mathematical equation. This algorithm uses data (information) provided as input and finds a pattern that determines whether the transaction is done by a fraud or a genuine customer. The historical data provided to the machine consist of previous transactions done by frauds and customers. In this case, the more frauds in the dataset, the better result.
4. Create Model- After training, the we have the model specific to the requirement of the business. In this case, if our fraud detection model is trained well, then it will be ready to detect the fraud and the genuine customer.

here, Since we are categorizing the entry in two categories- fraud or customer, therefore, this type of learning comes under Classification.

How can you tell the model is working?

After the training, to check that the model is working correctly, we show the model some data which it has never seen before, but which we know the fraud outcomes for. If the model detects the fraud correctly, we can deploy it to be used against the online business’ transactions. We also do some automatic common-sense analysis on recent date for which we do not have fraud labels to ensure the model will behave correctly when it is deployed.

There are certain suspicious situations which the model should always pick up on — some examples are:

1. High velocity of new payment methods eg. a customer adds new 10 payment cards in an hour.
2. Suspicious email address eg. a mismatch between the account name or name on the card, or rude/naughty words in the email.
3. A customer placing lots of orders of high value good eg. luxury alcohol.
4. Orders from a particularly suspicious location, shipping to a known fraud hotspot or a PO Box rather than a residential address.

Conclusion- With the involvement of Machine Learning in the field of money transfer, The rate of fraud should be minimized which actually happened in real world. It is really difficult for humans to keep track of every transaction done by the customers and determine whether it is a fraud or not. Now a days, upi payments, net banking, smart banking apps, credit/debit card transactions etc generates huge amount of data which needs to be analyzed in an attempt to minimize the fraud. However, it is still not up to the mark.