Credit Card

ML - randomForest

The results from TPOT point to using a Decision Tree algorithm.

Once we've selected our algorithm:

• Train a randomForest model in R.

• Deploy your model.

• Predict fraudulent credit card transactions.

The model that will be used is randomForest.

randomForest

Train the Model

Train the randomForest algorithm with the same dataset.

main_job.kjb

1. In Spoon, open the following main job:

/home/pentaho/Workshop--Data-Integration/Labs/Module 7 - Workflows/Machine Learning/Credit Card Fraud/solution/jb_fraud_main_job.kjb

2. Right-click on the train_model transformation and select Open Referenced Object -> Transformation.

train model

---

R Script Executor

1. Double-click on the rscrpt-train_model_randomForest step to bring up the configuration settings.

2. Under the Configure tab, ensure the Input Frames points to the step name sv-convert_booleans_to_numbers and the R Frame name: train.

R Script executor step

3. Set Row Handling to Number of Rows to Process: All.

4. Select the R script tab. Copy and paste the code snippets based on the Comments.

R script train

library(randomForest)

train.df <- as.data.frame(train)
rf <- randomForest(train.df$reported_as_fraud_historic ~ ., train.df, ntree=8, importance=TRUE)
save(rf, file="/home/pentaho/rf.rdata")

ok <- "Finished"
ok.df <- as.data.frame(ok)
ok.df

Predict Fraud

Using the 'trained' model - predict fraudulent credit activity.

main_job.kjb

1. In Spoon, open the following main job:

/home/pentaho/Workshop--Data-Integration/Labs/Module 7 - Workflows/Machine Learning/Credit Card Fraud/solution/jb_fraud_main_job.kjb

2. Right-click on the train_model transformation and select Open Referenced Object -> Transformation.

predict fraud

---

R Script Executor

1. Double-click on the rscrpt-predict step to bring up the configuration settings.

2. Under the Configure tab, ensure the Input Frames points to the step name sv-convert_booleans_to_numbers and the R Frame name: test.

R Script executor step

3. Set Row Handling to Number of Rows to Process: All.

4. Select the R script tab. Copy and paste the code snippets based on the Comments.

R script - predict

library(randomForest)

test.df <- as.data.frame(test)
load(file="/home/pentaho/rf.rdata")
pred <- predict(rf, newdata = test.df)
pred.df <- as.data.frame(pred)

submission <- data.frame(cbind(test.df,pred.df))
submission

Results

A spreadsheet formula is used to calculate %which can be used to trigger various actions.

1. Navigate to:

/home/pentaho/Workshop--Data-Integration/Labs/Module 7 - Workflows/Machine Learning/Credit Card Fraud/solution/output/credit_card_predict.xlsx

Fraud prediction