Icon

Projeto2

Keras Autoencoder for Fraud Detection - Training

This workflow trains an autoendcoder model to detect fraudulent transactions.

Keras Autoencoder Architecture
Data Preprocessing
Training the Autoencoder
Optimizing threshold K
Final Performance
Data Cleaning and Preparation

2/3 of negatives for training
Table Partitioner
1/3 of negativesand all positivesfor validation
Concatenate
Min-max normalization
Normalizer
Variable to Table Row
Table Writer
Normalizer (Apply)
Shape: 20
Keras Input Layer
Missing Value
Units: 16Activation: Sigmoid
Keras Dense Layer
Date&Time Part Extractor
Units: 29Activation: Sigmoid
Keras Dense Layer
String to Date&Time
Units: 19Activation: Sigmoid
Keras Dense Layer
Rule Engine
Units: 8Activation: Sigmoid
Keras Dense Layer
Rule Engine
10 % for validation
Table Partitioner
Units: 20Activation: Sigmoid
Keras Dense Layer
Bar Chart
Units: 29Activation: Sigmoid
Keras Dense Layer
One to Many
Math Formula
Train with Loss function=MSE Optimizer=Adam
Keras Network Learner
Write model
Keras Network Writer
Rule Engine
Scorer (JavaScript)
Counter Generation
Column Renamer
GroupBy
Column Filter
Apply network
Keras Network Executor
Rule Engine
Box Plot
Joiner
String Manipulation
Classifytransactions based onthreshold
Rule Engine
Math Formula
Class
Number to String
Rule Engine
Excel Reader
Top:Class = 0
Row Splitter
String to Number
Graphs
Threshold Optimization
Numeric Scorer
Normalizer (Apply)
Normalizerfor deployment
Model Writer
String to Number

Nodes

Extensions

Links

Download

To use this workflow in KNIME, download it from the below URL and open it in KNIME:

Download Workflow
Created by: Maarit Laukkanen
Created at: 2019-02-25
On NodePit since: 2026-05-14
Last update: 2026-05-14
Created with KNIME version: v5.11.0
Tags: autoencoderKerasneural networkdeep learningencoderdecoderreconstructioncredit card datafraud detectionanomaly detectionmodel evaluationbankingcredit cardfraudcybersecurity

Deploy, schedule, execute, and monitor your KNIME workflows locally, in the cloud or on-premises – with our brand new NodePit Runner.

Try NodePit Runner!