Today, we’re gonna build a model to predict fraudulent transactions in a bank.

We’ll start by:

1. Reviewing the dataset
2. We’ll then build a Machine Learning model for predicting fraud transactions.
3. And instantly see how we can make predictions and share them with others on the team.

So let’s get started.

Number 1: Reviewing the dataset

In this dataset, every single row is an individual transaction. And every column tells us something about that transaction. Like type, amount, balance, etc.  At the end, we have a column here called is Fraud. This tells us if a transaction in that row was flagged as fraudulent. Yes or No.  This is a historical dataset, and we now wanna use it to predict whether a new transaction is likely to be fraudulent.

Number 2: Building the machine learning model

To build a model, we start by uploading this CSV.  We can upload spreadsheets, Connect to apps & services like Salesforce and Dropbox And even databases like MySQL and BigQuery. Once uploaded, all you have to do is select your prediction column.  In this case, it’s the Is Fraud column.

From here, Obviously AI will automatically build a machine learning model to make predictions, in less than 1 minute.

Number 3: Make predictions and share with your team

Once your model is built, it’s ready to be used right away. No deploying, no maintenance. You can see it’s accuracy and performance details here.  Seems like we’re good to go! We start by heading to Personas to make a first prediction.

Let’s say we have a new transaction. This is a cash out transaction, with a total amount of $25,000 dollars. As I change these details, I can see the prediction for this transaction come up right here. This transaction has a 92% probability of not being fraudulent! Now, let’s head to export predictions. We can upload a list of new customers and get a prediction on each one. An API can also be used to automate these predictions and embed them in your own app or website. You will also have the option to get a shareable link to send to anyone on your team. It will enable them to use the model you built to start making their own predictions.

So, what would you like to predict today?