Fraud Detection on Financial Transactions with Machine Learning on Google Cloud | GSP774

Описание: Overview
In this lab you will explore the financial transactions data for fraud analysis, apply feature engineering and machine learning techniques to detect fraudulent activities using BigQuery ML.

A public financial transactions data will be used. The data contains the following columns:

type of the transaction
amount transferred
account id of origin and destination
new and old balances
relative time of transaction (number of hours from the start of the 30-day period)
isfraud flag
The target column isfraud includes the labels for the fraudulent transactions. Using these labels you will train supervised models for fraud detection and apply unsupervised models to detect anomalies.

In your own environment you can use the same data and follow the same steps to create models in AutoML Tables and/or Cloud AI Platform using Notebooks as an alternative, if you are familiar with those products.

The data for this lab is from the Kaggle site. If you do not have a Kaggle account, it's free to create one.

What you'll do:

Load data into BigQuery and explore.
Create new features in BigQuery.
Build an unsupervised model for anomaly detection.
Build supervised models (with logistic regression and boosted tree) for fraud detection.
Evaluate and compare the models and select the champion.
Use the selected model to predict the likelihood of fraud on a test data.
In this lab, you will use the BigQuery interface for feature engineering, model development, evaluation and prediction.

Participants that prefer Notebooks as the model development interface may choose to build models in AI Platform Notebooks instead of BigQuery ML. Then at the end of the lab, you can also complete the optional section. You can import open source libraries and create custom models or you can call BigQuery ML models within Notebooks using BigQuery magic commands.

If you want to train models in an automated way without any coding, you can use Google Cloud AutoML which builds models using state-of-the-art algorithms. The training process for AutoML would take almost 2 hours, that's why it is recommended to initiate it at the beginning of the lab, as soon as the data is prepared, so that you can see the results at the end. Check for the "Attention" phrase at the end of the data preparation step.

#gcp #googlecloud #qwiklabs #learn2earn #opensource