DataRobot API resources > R modeling workflow overview

R modeling workflow overview¶

This code example outlines how to use DataRobot's R client to train and experiment with models. It also offers ideas for integrating DataRobot with other products via the API.

Specifically, you will:

Create a project and run Autopilot.
Experiment with feature lists, modeling algorithms, and hyperparameters.
Choose the best model.
Perform an in-depth evaluation of the selected model.
Deploy a model into production in a few lines of code. This code example outlines how to use DataRobot's R client to train and experiment with models. It also offers ideas for integrating DataRobot with other products via the API.

Data used for this example¶

This walkthrough uses a synthetic dataset that illustrates a credit card company’s anti-money laundering (AML) compliance program, with the intent of detecting the following money-laundering scenarios:

A customer spends on the card, but overpays their credit card bill and seeks a cash refund for the difference.
A customer receives credits from a merchant without offsetting transactions, and either spends the money or requests a cash refund from the bank.

A rule-based engine is in place to produce an alert when it detects potentially suspicious activity consistent with the scenarios above. The engine triggers an alert whenever a customer requests a refund of any amount. Small refund requests are included because they could be a money launderer’s way of testing the refund mechanism or trying to establish refund requests as a normal pattern for their account.

The target feature is SAR, suspicious activity reports. It indicates whether or not the alert resulted in an SAR after manual review by investigators, which means that this project is a binary classification problem. The unit of analysis is an individual alert, so the model will be built on the alert level. Each alert will get a score ranging from 0 to 1, indicating the probability of being an alert leading to an SAR. The data consists of a mixture of numeric, categorical, and text data.

Setup¶

The following sections outline the required libraries and connection methods to begin building models with Autopilot.

Import Libraries¶

In [5]:

Copied!

library(dplyr)
library(ggplot2)
library(datarobot)
library(dplyr)
library(ggplot2)
library(datarobot)

Connect to DataRobot¶

To use R with DataRobot you first need to establish a connection between your machine and the DataRobot instance. Read more about different options for connecting to DataRobot from the client.

In [ ]:

Copied!

datarobot::ConnectToDataRobot(configPath = '<file-path-to-drconfig.yaml>')
datarobot::ConnectToDataRobot(configPath = '')

Upload a dataset¶

In [46]:

Copied!





# Set to the location of the training data via a local file or URL 
# Sample file location: '/Users/myuser/Downloads/DR_Demo_AML_Alert.csv'
dataset_file_path <- "https://s3.amazonaws.com/datarobot_public_datasets/DR_Demo_AML_Alert.csv"
training_data <- utils::read.csv(dataset_file_path)
test_data <- training_data[-c(2)]
head(training_data)
# Set to the location of the training data via a local file or URL 
# Sample file location: '/Users/myuser/Downloads/DR_Demo_AML_Alert.csv'
dataset_file_path <- "https://s3.amazonaws.com/datarobot_public_datasets/DR_Demo_AML_Alert.csv"
training_data <- utils::read.csv(dataset_file_path)
test_data <- training_data[-c(2)]
head(training_data)

A data.frame: 6 × 31
	ALERT	SAR	kycRiskScore	income	tenureMonths	creditScore	state	nbrPurchases90d	avgTxnSize90d	totalSpend90d	⋯	indCustReqRefund90d	totalRefundsToCust90d	nbrPaymentsCashLike90d	maxRevolveLine	indOwnsHome	nbrInquiries1y	nbrCollections3y	nbrWebLogins90d	nbrPointRed90d	PEP
	<int>	<int>	<int>	<int>	<int>	<int>	<chr>	<int>	<dbl>	<dbl>	⋯	<int>	<dbl>	<int>	<int>	<int>	<int>	<int>	<int>	<int>	<int>
1	1	0	3	110300	5	757	PA	10	153.80	1538.00	⋯	1	45.82	5	6000	0	3	0	6	1	0
2	1	0	2	107800	6	715	NY	22	1.59	34.98	⋯	1	67.40	0	10000	1	3	0	87	0	0
3	1	0	1	74000	13	751	MA	7	57.64	403.48	⋯	1	450.69	0	10000	0	3	0	6	0	0
4	1	0	0	57700	1	659	NJ	14	29.52	413.28	⋯	1	71.43	0	8000	1	5	0	7	2	0
5	1	0	1	59800	3	709	PA	54	115.77	6251.58	⋯	1	2731.39	3	7000	1	1	0	8	1	0
6	1	0	1	43500	11	717	CT	18	36.11	649.98	⋯	1	26.66	0	9000	0	2	1	3	2	0

Create a project and train models with Autopilot¶

In [ ]:

Copied!





# Create a project by uploading data. This will take a few minutes.
project <- SetupProject(dataSource = training_data, projectName = "SAR Detection", maxWait = 60 * 60)

# Set the project target and initiate Autopilot
SetTarget(project,
  target = "SAR")

# Block execution until Autopilot is complete
WaitForAutopilot(project)
# The `WaitForAutopilot()` function forces the R Kernel to wait until DataRobot has finished modeling before executing the next series of commands.

# Open the project's Leaderboard to monitor the progress in UI.
ViewWebProject("project-id") # Provide the project ID
# Create a project by uploading data. This will take a few minutes.
project <- SetupProject(dataSource = training_data, projectName = "SAR Detection", maxWait = 60 * 60)

# Set the project target and initiate Autopilot
SetTarget(project,
  target = "SAR")

# Block execution until Autopilot is complete
WaitForAutopilot(project)
# The `WaitForAutopilot()` function forces the R Kernel to wait until DataRobot has finished modeling before executing the next series of commands.

# Open the project's Leaderboard to monitor the progress in UI.
ViewWebProject("project-id") # Provide the project ID

Retrieve and review results from the Leaderboard¶

In [ ]:

Copied!

# Use the `ListModels()` function to retrieve a list of all the trained DataRobot models for a specified project.
ListModels(project)
# Use the `ListModels()` function to retrieve a list of all the trained DataRobot models for a specified project.
ListModels(project)

In [ ]:

Copied!

# Retrive the model DataRobot recommends for deployment
model <- GetRecommendedModel(project, type = RecommendedModelType$RecommendedForDeployment)

# Get a model's blueprint
GetModelBlueprintChart(project, model$modelId)
# Retrive the model DataRobot recommends for deployment
model <- GetRecommendedModel(project, type = RecommendedModelType$RecommendedForDeployment)

# Get a model's blueprint
GetModelBlueprintChart(project, model$modelId)

Experiment to get better results¶

When you run a project using Autopilot, DataRobot first creates blueprints based on the characteristics of your data and puts them in the Repository. Then, it chooses a subset from these to train; when training completes, these are the blueprints you’ll find on the Leaderboard. After the Leaderboard is populated, it can be useful to train some of those blueprints that DataRobot skipped. For example, you can try a more complex Keras blueprint like Keras Residual AutoInt Classifier using Training Schedule (3 Attention Layers with 2 Heads, 2 Layers: 100, 100 Units). In some cases, you may want to directly access the trained model through R and retrain it with a different feature list or tune its hyperparameters.

Find blueprints trained for the project from the Leaderboard¶

In [ ]:

Copied!

modelsInLeaderboard <- ListModels(project)
modelsInLeaderboard_df <- as.data.frame(modelsInLeaderboard)
modelsInLeaderboard <- ListModels(project)
modelsInLeaderboard_df <- as.data.frame(modelsInLeaderboard)

Train a model using a different feature list¶

In [ ]:

Copied!





for (i in 1:length(modelsInLeaderboard_df)){
    job <- RequestNewModel(project, modelsInLeaderboard_df[[i]])
    WaitForJobToComplete(project, job, maxWait=600)
}
for (i in 1:length(modelsInLeaderboard_df)){
    job <- RequestNewModel(project, modelsInLeaderboard_df[[i]])
    WaitForJobToComplete(project, job, maxWait=600)
}

Tune hyperparameters for a model¶

In [ ]:

Copied!

StartTuningSession(model)
StartTuningSession(model)

Select the top-performing model¶

In [ ]:

Copied!

# Use the `ListModels()` function to retrieve a list of all the trained DataRobot models for a specified project
ListModels(project)

# Retrieve the model DataRobot recommends for deployment
model <- GetRecommendedModel(project, type = RecommendedModelType$RecommendedForDeployment)
# Use the `ListModels()` function to retrieve a list of all the trained DataRobot models for a specified project
ListModels(project)

# Retrieve the model DataRobot recommends for deployment
model <- GetRecommendedModel(project, type = RecommendedModelType$RecommendedForDeployment)

Calculate Feature Impact¶

In [19]:

Copied!





# Select a model from the Leaderboard:
model <- GetModel(project, model$modelId)

# Retrieve the top 10 most impactful features:
feature_impact <- GetFeatureImpact(model)
feature_impact <- feature_impact[order(-feature_impact$impactNormalized), ] %>% slice(1:10)
feature_impact
# Select a model from the Leaderboard:
model <- GetModel(project, model$modelId)

# Retrieve the top 10 most impactful features:
feature_impact <- GetFeatureImpact(model)
feature_impact <- feature_impact[order(-feature_impact$impactNormalized), ] %>% slice(1:10)
feature_impact

A data.frame: 10 × 4
featureName	impactNormalized	impactUnnormalized	redundantWith
<chr>	<dbl>	<dbl>	<lgl>
totalMerchCred90d	1.00000000	0.149085915	NA
nbrCustReqRefunds90d	0.51163007	0.076276837	NA
totalRefundsToCust90d	0.42568163	0.063463135	NA
totalSpend90d	0.14489628	0.021601995	NA
totalPaymentAmt90d	0.12352879	0.018416403	NA
avgTxnSize90d	0.06139331	0.009152878	NA
overpaymentInd90d	0.05853019	0.008726026	NA
csrNotes	0.05485649	0.008178330	NA
income	0.03974456	0.005925353	NA
nbrPaymentsCashLike90d	0.03641069	0.005428321	NA

Train a model with a new feature list¶

In [ ]:

Copied!





# Create a feature list using the top 10 features:
top10_feat_list <- CreateFeaturelist(project, "Top 10 Features Across Models",
                                     feature_impact$featureName)

# Retrain models using the new feature list:
StartNewAutoPilot(project, top10_feat_list$featurelistId, 
                  mode = AutopilotMode$Quick)
# Create a feature list using the top 10 features:
top10_feat_list <- CreateFeaturelist(project, "Top 10 Features Across Models",
                                     feature_impact$featureName)

# Retrain models using the new feature list:
StartNewAutoPilot(project, top10_feat_list$featurelistId, 
                  mode = AutopilotMode$Quick)

Select the best-performing model¶

In [ ]:

Copied!

model <- ListModels(project)[[1]]
model
model <- ListModels(project)[[1]]
model

In-depth model evaluation¶

Retrieve and plot the ROC curve¶

In [24]:

Copied!





roc <- GetRocCurve(model,
                   source = 'validation')
roc_df <- roc$rocPoints
head(roc_df)
roc <- GetRocCurve(model,
                   source = 'validation')
roc_df <- roc$rocPoints
head(roc_df)

A data.frame: 6 × 17
	accuracy	f1Score	falseNegativeScore	trueNegativeScore	truePositiveScore	falsePositiveScore	trueNegativeRate	falsePositiveRate	truePositiveRate	matthewsCorrelationCoefficient	positivePredictiveValue	negativePredictiveValue	threshold	fractionPredictedAsPositive	fractionPredictedAsNegative	liftPositive	liftNegative
	<dbl>	<dbl>	<int>	<int>	<int>	<int>	<dbl>	<dbl>	<dbl>	<dbl>	<dbl>	<dbl>	<dbl>	<dbl>	<dbl>	<dbl>	<dbl>
1	0.897500	0.00000000	164	1436	0	0	1.0000000	0.0000000000	0.000000000	0.00000000	0.0000000	0.8975000	1.0000000	0.000000	1.000000	0.000000	1.000000
2	0.898125	0.01212121	163	1436	1	0	1.0000000	0.0000000000	0.006097561	0.07399989	1.0000000	0.8980613	0.8840490	0.000625	0.999375	9.756098	1.000625
3	0.900625	0.07017544	158	1435	6	1	0.9993036	0.0006963788	0.036585366	0.16493143	0.8571429	0.9008161	0.8461373	0.004375	0.995625	8.362369	1.003695
4	0.903750	0.12500000	153	1435	11	1	0.9993036	0.0006963788	0.067073171	0.23334555	0.9166667	0.9036524	0.8384385	0.007500	0.992500	8.943089	1.006855
5	0.904375	0.14525140	151	1434	13	2	0.9986072	0.0013927577	0.079268293	0.24509809	0.8666667	0.9047319	0.8311916	0.009375	0.990625	8.455285	1.008058
6	0.905000	0.18279570	147	1431	17	5	0.9965181	0.0034818942	0.103658537	0.26091604	0.7727273	0.9068441	0.8193629	0.013750	0.986250	7.538803	1.010411

In [25]:

Copied!





dr_roc_green <- '#03c75f'
ggplot(data = roc_df,
       mapping = aes(x = falsePositiveRate,
                     y = truePositiveRate)) +
    geom_segment(mapping = aes(x = 0, xend = 1, y = 0, yend = 1), color = "white") +
    geom_point(color = dr_roc_green) +
    xlab("False Positive Rate (Fallout)") + 
    ylab("True Positive Rate (Sensitivity)") + 
    theme_dark()
dr_roc_green <- '#03c75f'
ggplot(data = roc_df,
       mapping = aes(x = falsePositiveRate,
                     y = truePositiveRate)) +
    geom_segment(mapping = aes(x = 0, xend = 1, y = 0, yend = 1), color = "white") +
    geom_point(color = dr_roc_green) +
    xlab("False Positive Rate (Fallout)") + 
    ylab("True Positive Rate (Sensitivity)") + 
    theme_dark()

Retrieve and plot Feature Impact¶

In [20]:

Copied!





# Retrieve the top 10 most impactful features:
feature_impact <- GetFeatureImpact(model)
feature_impact <- feature_impact[order(-feature_impact$impactNormalized),] %>% slice(1:10)

# Create plot of top 10 features based on Feature Impact
ggplot(data = feature_impact,
       mapping = aes(
           x = featureName, 
           y = impactNormalized)) + 
    geom_col() +
    labs(x = "Feature")
# Retrieve the top 10 most impactful features:
feature_impact <- GetFeatureImpact(model)
feature_impact <- feature_impact[order(-feature_impact$impactNormalized),] %>% slice(1:10)

# Create plot of top 10 features based on Feature Impact
ggplot(data = feature_impact,
       mapping = aes(
           x = featureName, 
           y = impactNormalized)) + 
    geom_col() +
    labs(x = "Feature")

Score data before deployment¶

In [ ]:

Copied!

test_data <- training_data[ -c(2) ]
head(test_data)
test_data <- training_data[ -c(2) ]
head(test_data)

In [ ]:

Copied!





# Uploading the testing dataset
scoring <- UploadPredictionDataset(project, dataSource = test_data)

# Requesting prediction
predict_job_id <- RequestPredictions(project, modelId = model$modelId, datasetId = scoring$id)

# Grabbing predictions
predictions_prob <- GetPredictions(project, predictId = predict_job_id, type = "probability")
head(predictions_prob)
# Uploading the testing dataset
scoring <- UploadPredictionDataset(project, dataSource = test_data)

# Requesting prediction
predict_job_id <- RequestPredictions(project, modelId = model$modelId, datasetId = scoring$id)

# Grabbing predictions
predictions_prob <- GetPredictions(project, predictId = predict_job_id, type = "probability")
head(predictions_prob)

Compute Prediction Explanations¶

In [ ]:

Copied!

GetPredictionExplanations(model, test_data)
GetPredictionExplanations(model, test_data)

Deploy a model¶

After identifying the best-performing models, you can deploy them and use DataRobot's REST API to make HTTP requests and return predictions. You can also configure batch prediction jobs to write back into your environment of choice.

Once deployed, access monitoring capabilities such as:

In [ ]:

Copied!





prediction_server <- ListPredictionServers()[[1]]

deployment <- CreateDeployment(model, 
                               label = 'New Deployment', 
                               description = 'A new deployment',
                               defaultPredictionServerId = prediction_server$id)
deployment
prediction_server <- ListPredictionServers()[[1]]

deployment <- CreateDeployment(model, 
                               label = 'New Deployment', 
                               description = 'A new deployment',
                               defaultPredictionServerId = prediction_server$id)
deployment

R modeling workflow overview¶

Data used for this example¶

Setup¶

Import Libraries¶

Connect to DataRobot¶

Upload a dataset¶

Create a project and train models with Autopilot¶

Retrieve and review results from the Leaderboard¶

Experiment to get better results¶

Find blueprints trained for the project from the Leaderboard¶

Train a model using a different feature list¶

Tune hyperparameters for a model¶

Select the top-performing model¶

Calculate Feature Impact¶

Train a model with a new feature list¶

Select the best-performing model¶

In-depth model evaluation¶

Retrieve and plot the ROC curve¶

Retrieve and plot Feature Impact¶

Score data before deployment¶

Compute Prediction Explanations¶

Deploy a model¶

Was this page helpful?

Great! Let us know what you found helpful.

What can we do to improve the content?