This document will walk through you how to deploy a credit risk model as a web service, using the mrsdeploy package that ships with Microsoft R Client and R Server.
It will start by creating the model locally, then publish it as a web service, and then share it with other authenticated users for consumption.
+The mrsdeploy package, delivered with Microsoft R Client and R Server, provides functions for:
1 Establishing a remote session in a R console application for the purposes of executing code on that server
+2 Publishing and managing an R web service that is backed by the R code block or script you provided.
+Each feature can be used independently, but the greatest value is achieved when you can leverage both.
+This document will walk through you how to deploy a credit risk model as a web service, using the mrsdeploy package.
It will start by modelling locally, then publish it as a web service, and then share it with other authenticated users for consumption, and finally manage and update the web service.
## Not adding a normalizer.
-## Making per-feature arrays
-## Changing data from row-wise to column-wise
-## Beginning processing data.
+ minSplit=10,
+ unbalancedSets=FALSE,
+ verbose=0)## Beginning processing data.
## Rows Read: 128977, Read Time: 0, Transform Time: 0
-## Beginning processing data.
-## Processed 128977 instances
-## Binning and forming Feature objects
-## Reserved memory for tree learner: 219024 bytes
-## Starting to train ...
-## Not training a calibrator because it is not needed.
-## Elapsed time: 00:00:02.1784776summary(model_rxtrees)## Elapsed time: 00:00:00.2568312model_rxtrees## Call:
## rxFastTrees(formula = form, data = data[train, c(target, vars)],
## type = "binary", numTrees = 100, numLeaves = 20, learningRate = 0.2,
-## minSplit = 10)
+## minSplit = 10, unbalancedSets = FALSE, verbose = 0)
##
-## Predictor for: bad_flag~amount_6+pur_6+avg_pur_amt_6+avg_interval_pur_6+credit_limit+age+income+sex+education+marital_status
-## Data: data[train, c(target, vars)]
-##
-## Per-feature gain:
-## credit_limit: 1
-## age: 0.8760216
-## education.undergraduate: 0.3149455
-## education.polytechnics: 0.3069959
-## education.master: 0.2611516
-## income: 0.2581172
-## sex.female: 0.2488511
-## marital_status.single: 0.2456198
-## marital_status.married: 0.2349416
-## education.high_school: 0.2142461
-## sex.male: 0.1726051
-## amount_6: 0.1578717
-## education.middle_school: 0.1311938
-## avg_pur_amt_6: 0.08475202
-## avg_interval_pur_6: 0.02046135# Produce a prediction function that can use the model
creditRiskPrediction <- function(account_id, amount_6, pur_6, avg_pur_amt_6, avg_interval_pur_6,
credit_limit, marital_status, sex, education, income, age)
-{
- staticdata <- data.frame(account_id=c("a_1055521381828530", "a_1055521125532160", "a_914800337488587", "a_844428146542225", "a_844428047550192"),
- amount_6=c(106.99, 212.49, 118.00, 151.19, 148.39),
- pur_6=c(1, 1, 1, 1, 1),
- avg_pur_amt_6=c(106.99, 212.49, 118.00, 151.19, 148.39),
- avg_interval_pur_6=c(0, 0, 0, 0, 0),
- credit_limit=c(2.21, -0.96, 3.82, -0.39, -0.45),
- marital_status=c("single", "married", "single", "married", "single"),
- sex=c("female", "male", "male", "female", "female"),
- education=c("master", "high_school", "middle_school",
- "polytechnics", "undergraduate"),
- income=c(0, 6.27, 17.15, 9.94, 19.93),
- age=c(33, 47, 30, 51, 32))
-
- inputdata <- data.frame(account_id=account_id,
+{
+ newdata <- data.frame(account_id=account_id,
amount_6=amount_6,
pur_6=pur_6,
avg_pur_amt_6=avg_pur_amt_6,
@@ -305,13 +271,10 @@ creditRiskPrediction <- function(account_id, amount_6, pur_6, avg_pur_amt_6,
income=income,
age=age)
- newdata <- rbind(inputdata, staticdata)
-
- pred <- rxPredict(model_rxtrees, data=newdata)[, c(1, 3)]
+ pred <- rxPredict(modelObject=model_rxtrees, data=newdata)[, c(1, 3)]
pred <- cbind(newdata$account_id, pred)
names(pred) <- c("account_id", "scored_label", "scored_prob")
- pred <- pred[1, ]
- pred
+ pred
}
# Test function locally by printing results
@@ -328,10 +291,10 @@ pred <- creditRiskPrediction(account_id="a_1055521029582310",
income=12.36,
age=38)## Beginning processing data.
-## Rows Read: 6, Read Time: 0, Transform Time: 0
+## Rows Read: 1, Read Time: 0.001, Transform Time: 0
## Beginning processing data.
-## Elapsed time: 00:00:00.1521922
-## Finished writing 6 rows.
+## Elapsed time: 00:00:00.3347313
+## Finished writing 1 rows.
## Writing completed.print(pred)## account_id scored_label scored_prob
@@ -376,7 +339,7 @@ api <- publishService(
Finally, we can consume the service in R directly after publishing it to verify that the results are as expected.
+After publishing it , we can consume the service in R directly to verify that the results are as expected.
# Get service and assign service to the variable `api`.
api <- getService("crpService", "v1.0.0")
@@ -400,6 +363,52 @@ result <- api$creditRiskPrediction(account_id="a_1055521029582310",
print(result$output("pred")) ## account_id scored_label scored_prob
## 1 a_1055521029582310 no 0.0250131245702505In the process of production, we could manage and update the web service timely.
+# Load the pre-trained optimal model obtained from the template of CreditRiskScale.
+
+load(file="model_rxtrees.RData")
+
+model_rxtrees## Call:
+## rxFastTrees(formula = form, data = data_split[[1]], type = "binary",
+## numTrees = 500, numLeaves = 20, learningRate = 0.3, minSplit = 10,
+## unbalancedSets = FALSE, verbose = 0)
+##
+## FastTreeBinaryClassification (BinaryClassifierTrainer) for: bad_flag~amount_6+pur_6+avg_pur_amt_6+avg_interval_pur_6+credit_limit+age+income+sex+education+marital_status
+## Data: data_split[[1]] (RxXdfData Data Source)
+## File name:
+## C:\Demo\acceleratoRs-master\CreditRiskPrediction\Code\file4b9418067c23..train.1.xdfapi <- updateService(name="crpService",
+ v="v1.0.0",
+ model=model_rxtrees,
+ descr="Update the model hyper-parameters")
+
+# Re-test the updated service by consuming it
+
+result <- api$creditRiskPrediction(account_id="a_1055521029582310",
+ amount_6=173.22,
+ pur_6=1,
+ avg_pur_amt_6=173.22,
+ avg_interval_pur_6=0,
+ credit_limit=5.26,
+ marital_status="married",
+ sex="male",
+ education="undergraduate",
+ income=12.36,
+ age=38)
+
+# Print response output named `answer`
+
+print(result$output("pred")) ## account_id scored_label scored_prob
+## 1 a_1055521029582310 no 0.0156644005328417Last but not least, we can get the json file that is needed for application integration.
# Get this service's `swagger.json` file that is needed for web application integration
swagger <- api$swagger(json = FALSE)
diff --git a/CreditRiskPrediction/Code/CreditRiskDeploy.ipynb b/CreditRiskPrediction/Code/CreditRiskDeploy.ipynb
deleted file mode 100644
index 3e59116..0000000
--- a/CreditRiskPrediction/Code/CreditRiskDeploy.ipynb
+++ /dev/null
@@ -1,449 +0,0 @@
-{
- "cells": [
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "---\n",
- "title: \"Deploy a Credit Risk Model as a Web Service\"\n",
- "author: \"Fang Zhou, Data Scientist, Microsoft\"\n",
- "date: \"`r Sys.Date()`\"\n",
- "output: html_document\n",
- "---"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {
- "attributes": {
- "classes": [],
- "id": "",
- "include": "FALSE,",
- "purl": "FALSE"
- }
- },
- "outputs": [],
- "source": [
- "knitr::opts_chunk$set(echo = TRUE,\n",
- " fig.width = 8,\n",
- " fig.height = 5,\n",
- " fig.align='center',\n",
- " dev = \"png\")"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "## 1 Introduction\n",
- "\n",
- "This document will walk through you how to deploy a credit risk model as a web service,\n",
- "using the `mrsdeploy` package that ships with Microsoft R Client and R Server. \n",
- "\n",
- "It will start by creating the model locally, then publish it as a web service, and then share it \n",
- "with other authenticated users for consumption. \n",
- "\n",
- "## 2 Automated Credit Risk Model Deployment\n",
- "\n",
- "### 2.1 Setup\n",
- "\n",
- "We load the required R packages."
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {
- "attributes": {
- "classes": [],
- "error": "FALSE",
- "id": "",
- "message": "FALSE,",
- "warning": "FALSE,"
- }
- },
- "outputs": [],
- "source": [
- "## Setup\n",
- "\n",
- "# Load the required packages into the R session.\n",
- "\n",
- "library(rattle) # Use normVarNames().\n",
- "library(dplyr) # Wrangling: tbl_df(), group_by(), print(), glimpse().\n",
- "library(magrittr) # Pipe operator %>% %<>% %T>% equals().\n",
- "library(scales) # Include commas in numbers.\n",
- "library(MicrosoftML) # Build models using Microsoft ML algortihms.\n",
- "library(mrsdeploy) # Publish an R model as a web service."
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "Then, the dataset processedSimu is ingested for demonstration. This dataset was created by the data preprocessing steps in the data science accelerator for credit risk prediction."
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {
- "attributes": {
- "classes": [],
- "error": "FALSE",
- "id": "",
- "message": "FALSE,",
- "warning": "FALSE,"
- }
- },
- "outputs": [],
- "source": [
- "## Data Ingestion\n",
- "\n",
- "# Identify the source location of the dataset.\n",
- "\n",
- "#DATA <- \"../../Data/\"\n",
- "#txn_fname <- file.path(DATA, \"Raw/processedSimu.csv\")\n",
- "\n",
- "wd <- getwd()\n",
- "\n",
- "dpath <- \"../Data\"\n",
- "data_fname <- file.path(wd, dpath, \"processedSimu.csv\")\n",
- "\n",
- "# Ingest the dataset.\n",
- "\n",
- "data <- read.csv(file=data_fname) %T>% \n",
- " {dim(.) %>% comma() %>% cat(\"\\n\")}\n",
- "\n",
- "# A glimpse into the data.\n",
- "\n",
- "glimpse(data)"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "### 2.2 Model Locally\n",
- "\n",
- "Now, let's get started to build an R model based web service. \n",
- "\n",
- "First of all, we create a machine learning fast tree model on the dataset processedSimu by using the function `rxFastTrees()` from the `MicrosoftML` package. This model could be used to predict whether an account will default or to predict its probability of default, given some transaction statistics and demographic & bank account information as inputs."
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {
- "attributes": {
- "classes": [],
- "error": "FALSE",
- "id": "",
- "message": "FALSE,",
- "warning": "FALSE,"
- }
- },
- "outputs": [],
- "source": [
- "## Variable roles.\n",
- "\n",
- "# Target variable\n",
- "\n",
- "target <- \"bad_flag\"\n",
- "\n",
- "# Note any identifier.\n",
- "\n",
- "id <- c(\"account_id\") %T>% print() \n",
- "\n",
- "# Note the available variables as model inputs.\n",
- "\n",
- "vars <- setdiff(names(data), c(target, id))"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {
- "attributes": {
- "classes": [],
- "error": "FALSE",
- "id": "",
- "message": "FALSE,",
- "warning": "FALSE,"
- }
- },
- "outputs": [],
- "source": [
- "# Split Data\n",
- "\n",
- "set.seed(42)\n",
- "\n",
- "data <- data[order(runif(nrow(data))), ]\n",
- "\n",
- "train <- sample(nrow(data), 0.70 * nrow(data))\n",
- "test <- setdiff(seq_len(nrow(data)), train)"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {
- "attributes": {
- "classes": [],
- "error": "FALSE",
- "id": "",
- "message": "FALSE,",
- "warning": "FALSE,"
- }
- },
- "outputs": [],
- "source": [
- "# Prepare the formula\n",
- "\n",
- "top_vars <- c(\"amount_6\", \"pur_6\", \"avg_pur_amt_6\", \"avg_interval_pur_6\", \"credit_limit\", \"age\", \"income\", \"sex\", \"education\", \"marital_status\")\n",
- "\n",
- "form <- as.formula(paste(target, paste(top_vars, collapse=\"+\"), sep=\"~\"))\n",
- "form"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {
- "attributes": {
- "classes": [],
- "error": "FALSE",
- "id": "",
- "message": "FALSE,",
- "warning": "FALSE,"
- }
- },
- "outputs": [],
- "source": [
- "# Train model: rxFastTrees\n",
- "\n",
- "model_rxtrees <- rxFastTrees(formula=form,\n",
- " data=data[train, c(target, vars)],\n",
- " type=\"binary\",\n",
- " numTrees=100,\n",
- " numLeaves=20,\n",
- " learningRate=0.2,\n",
- " minSplit=10)\n",
- "\n",
- "summary(model_rxtrees)"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {
- "attributes": {
- "classes": [],
- "error": "FALSE",
- "id": "",
- "message": "FALSE,",
- "warning": "FALSE,"
- }
- },
- "outputs": [],
- "source": [
- "# Produce a prediction function that can use the model\n",
- "\n",
- "creditRiskPrediction <- function(account_id, amount_6, pur_6, avg_pur_amt_6, avg_interval_pur_6, \n",
- " credit_limit, marital_status, sex, education, income, age)\n",
- "{\n",
- " staticdata <- data.frame(account_id=c(\"a_1055521381828530\", \"a_1055521125532160\", \"a_914800337488587\", \"a_844428146542225\", \"a_844428047550192\"),\n",
- " amount_6=c(106.99, 212.49, 118.00, 151.19, 148.39),\n",
- " pur_6=c(1, 1, 1, 1, 1),\n",
- " avg_pur_amt_6=c(106.99, 212.49, 118.00, 151.19, 148.39),\n",
- " avg_interval_pur_6=c(0, 0, 0, 0, 0),\n",
- " credit_limit=c(2.21, -0.96, 3.82, -0.39, -0.45),\n",
- " marital_status=c(\"single\", \"married\", \"single\", \"married\", \"single\"),\n",
- " sex=c(\"female\", \"male\", \"male\", \"female\", \"female\"),\n",
- " education=c(\"master\", \"high_school\", \"middle_school\", \n",
- " \"polytechnics\", \"undergraduate\"),\n",
- " income=c(0, 6.27, 17.15, 9.94, 19.93),\n",
- " age=c(33, 47, 30, 51, 32))\n",
- " \n",
- " inputdata <- data.frame(account_id=account_id,\n",
- " amount_6=amount_6, \n",
- " pur_6=pur_6, \n",
- " avg_pur_amt_6=avg_pur_amt_6, \n",
- " avg_interval_pur_6=avg_interval_pur_6, \n",
- " credit_limit=credit_limit, \n",
- " marital_status=marital_status, \n",
- " sex=sex, \n",
- " education=education, \n",
- " income=income, \n",
- " age=age)\n",
- " \n",
- " newdata <- rbind(inputdata, staticdata)\n",
- " \n",
- " pred <- rxPredict(model_rxtrees, data=newdata)[, c(1, 3)]\n",
- " pred <- cbind(newdata$account_id, pred)\n",
- " names(pred) <- c(\"account_id\", \"scored_label\", \"scored_prob\")\n",
- " pred <- pred[1, ]\n",
- " pred\n",
- "}\n",
- "\n",
- "# Test function locally by printing results\n",
- "\n",
- "pred <- creditRiskPrediction(account_id=\"a_1055521029582310\",\n",
- " amount_6=173.22, \n",
- " pur_6=1, \n",
- " avg_pur_amt_6=173.22, \n",
- " avg_interval_pur_6=0, \n",
- " credit_limit=5.26, \n",
- " marital_status=\"married\", \n",
- " sex=\"male\", \n",
- " education=\"undergraduate\", \n",
- " income=12.36, \n",
- " age=38)\n",
- "\n",
- "print(pred)"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "### 2.2 Publish model as a web service\n",
- "\n",
- "The second procedure is to publish the model as a web service by following the below steps.\n",
- "\n",
- "Step 1: From your local R IDE, log into Microsoft R Server with your credentials using the appropriate authentication function from the `mrsdeploy` package (remoteLogin or remoteLoginAAD). \n",
- "\n",
- "For simplicity, the code below uses the basic local admin account for authentication with the remoteLogin function and `session = false` so that no remote R session is started."
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {
- "attributes": {
- "classes": [],
- "error": "FALSE",
- "id": "",
- "message": "FALSE,",
- "warning": "FALSE,"
- }
- },
- "outputs": [],
- "source": [
- "# Use `remoteLogin` to authenticate with R Server using \n",
- "# the local admin account. Use session = false so no \n",
- "# remote R session started\n",
- "\n",
- "remoteLogin(\"http://localhost:12800\", \n",
- " username=\"admin\", \n",
- " password=\"P@ssw0rd\",\n",
- " session=FALSE)"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "Now, you are successfully connected to the remote R Server.\n",
- "\n",
- "Step 2: Publish the model as a web service to R Server using the `publishService()` function from the `mrsdeploy` package. \n",
- "\n",
- "In this example, you publish a web service called \"crpService\" using the model `model_rxtrees` and the function `creditRiskPrediction()`. As an input, the service takes a list of transaction statistics and demographic & bank account information represented as numerical or categorical. As an output, an R data frame including the account id, the predicted label of default, and the probability of default for the given individual account, has of being achieved with the pre-defined credit risk prediction function. \n",
- "\n",
- "When publishing, you must specify, among other parameters, a service name and version, the R code, the inputs, as well as the outputs that application developers will need to integrate in their applications."
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {
- "attributes": {
- "classes": [],
- "error": "FALSE",
- "id": "",
- "message": "FALSE,",
- "warning": "FALSE,"
- }
- },
- "outputs": [],
- "source": [
- "# Publish a web service\n",
- "\n",
- "api <- publishService(\n",
- " \"crpService\",\n",
- " code=creditRiskPrediction,\n",
- " model=model_rxtrees,\n",
- " inputs=list(account_id=\"character\",\n",
- " amount_6=\"numeric\", \n",
- " pur_6=\"numeric\", \n",
- " avg_pur_amt_6=\"numeric\", \n",
- " avg_interval_pur_6=\"numeric\", \n",
- " credit_limit=\"numeric\", \n",
- " marital_status=\"character\", \n",
- " sex=\"character\", \n",
- " education=\"character\", \n",
- " income=\"numeric\", \n",
- " age=\"numeric\"),\n",
- " outputs=list(pred=\"data.frame\"),\n",
- " v=\"v1.0.0\")"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "### 2.3 Test the service by consuming it in R\n",
- "\n",
- "Finally, we can consume the service in R directly after publishing it to verify that the results are as expected."
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {
- "attributes": {
- "classes": [],
- "error": "TRUE",
- "id": "",
- "message": "FALSE,",
- "warning": "FALSE,"
- }
- },
- "outputs": [],
- "source": [
- "# Get service and assign service to the variable `api`.\n",
- "\n",
- "api <- getService(\"crpService\", \"v1.0.0\")\n",
- "\n",
- "# Consume service by calling function, `creditRiskPrediction` contained in this service\n",
- "\n",
- "result <- api$creditRiskPrediction(account_id=\"a_1055521029582310\",\n",
- " amount_6=173.22, \n",
- " pur_6=1, \n",
- " avg_pur_amt_6=173.22, \n",
- " avg_interval_pur_6=0, \n",
- " credit_limit=5.26, \n",
- " marital_status=\"married\", \n",
- " sex=\"male\", \n",
- " education=\"undergraduate\", \n",
- " income=12.36, \n",
- " age=38)\n",
- "\n",
- "# Print response output named `answer`\n",
- "\n",
- "print(result$output(\"pred\")) \n",
- "\n",
- "# Get this service's `swagger.json` file that is needed for web application integration\n",
- "\n",
- "swagger <- api$swagger(json = FALSE)\n",
- "\n",
- "# Delete the service to make the script re-runable\n",
- "\n",
- "deleteService(name=\"crpService\", v=\"v1.0.0\")"
- ]
- }
- ],
- "metadata": {},
- "nbformat": 4,
- "nbformat_minor": 1
-}
diff --git a/CreditRiskPrediction/Code/CreditRiskScale.Rmd b/CreditRiskPrediction/Code/CreditRiskScale.Rmd
index 81aa1da..47fe16f 100644
--- a/CreditRiskPrediction/Code/CreditRiskScale.Rmd
+++ b/CreditRiskPrediction/Code/CreditRiskScale.Rmd
@@ -77,7 +77,9 @@ rxGetVarInfo(data)
Now, let's get started to build credit risk models by leveraging different machine learning algorithms from the `MicrosoftML` package.
-First of all, we create individual machine learning models on the dataset processedSimu by using the functions `rxLogisticRegression()`, `rxFastTrees()`, and `rxFastForest()`.
+First of all, we create individual machine learning models on the dataset processedSimu.xdf by using the functions `rxLogisticRegression()`, `rxFastForest()`, `rxFastTrees()`.
+
+From the credit risk prediction template, we know that gradient boosting is the most suitable algorithm for this example, considering the overall performance. Therefore, the models implemented by the function `rxFastTrees()` with different sets of parameters are trained respectively.
```{r, message=FALSE, warning=FALSE, error=FALSE}
## Variable roles.
@@ -100,9 +102,6 @@ vars <- setdiff(names(data), c(target, id))
set.seed(42)
-train <- sample(nrow(data), 0.70 * nrow(data))
-test <- setdiff(seq_len(nrow(data)), train)
-
# Add training/testing flag to each observation.
data %<>%
@@ -138,21 +137,9 @@ time_rxlogit <- system.time(
formula=form,
data=data_split[[1]],
type="binary",
- l1Weight=1)
-)
-
-# Train model: rxFastTrees
-
-time_rxtrees <- system.time(
-
- model_rxtrees <- rxFastTrees(
- formula=form,
- data=data_split[[1]],
- type="binary",
- numTrees=100,
- numLeaves=20,
- learningRate=0.2,
- minSplit=10)
+ l1Weight=1,
+ verbose=0
+ )
)
# Train model: rxFastForest
@@ -165,11 +152,75 @@ time_rxforest <- system.time(
type="binary",
numTrees=100,
numLeaves=20,
- minSplit=10)
+ minSplit=10,
+ verbose=0
+ )
+)
+
+# Train model: rxFastTrees
+
+time_rxtrees1 <- system.time(
+
+ model_rxtrees1 <- rxFastTrees(
+ formula=form,
+ data=data_split[[1]],
+ type="binary",
+ numTrees=100,
+ numLeaves=20,
+ learningRate=0.2,
+ minSplit=10,
+ unbalancedSets=FALSE,
+ verbose=0
+ )
+)
+
+time_rxtrees2 <- system.time(
+
+ model_rxtrees2 <- rxFastTrees(
+ formula=form,
+ data=data_split[[1]],
+ type="binary",
+ numTrees=500,
+ numLeaves=20,
+ learningRate=0.2,
+ minSplit=10,
+ unbalancedSets=FALSE,
+ verbose=0
+ )
+)
+
+time_rxtrees3 <- system.time(
+
+ model_rxtrees3 <- rxFastTrees(
+ formula=form,
+ data=data_split[[1]],
+ type="binary",
+ numTrees=500,
+ numLeaves=20,
+ learningRate=0.3,
+ minSplit=10,
+ unbalancedSets=FALSE,
+ verbose=0
+ )
+)
+
+time_rxtrees4 <- system.time(
+
+ model_rxtrees4 <- rxFastTrees(
+ formula=form,
+ data=data_split[[1]],
+ type="binary",
+ numTrees=500,
+ numLeaves=20,
+ learningRate=0.3,
+ minSplit=10,
+ unbalancedSets=TRUE,
+ verbose=0
+ )
)
```
-Next, we build an ensemble of models by using the function `rxEnsemble()`.
+Next, we build an ensemble of fast tree models by using the function `rxEnsemble()`.
```{r, message=FALSE, warning=FALSE, error=FALSE}
# Train an ensemble model.
@@ -180,20 +231,27 @@ time_ensemble <- system.time(
formula=form,
data=data_split[[1]],
type="binary",
- trainers=list(logisticRegression(), fastTrees(), fastForest()),
- replace=TRUE
+ trainers=list(fastTrees(),
+ fastTrees(numTrees=500),
+ fastTrees(numTrees=500, learningRate=0.3),
+ fastTrees(numTrees=500, learningRate=0.3, unbalancedSets=TRUE)),
+ combineMethod="vote",
+ replace=TRUE,
+ verbose=0
)
)
```
### 2.3 Model Evaluation
-Finally, we evaluate and compare the above built models.
+Finally, we evaluate and compare the above built models at various aspects.
```{r, message=FALSE, warning=FALSE, error=FALSE}
# Predict
-models <- list(model_rxlogit, model_rxtrees, model_rxforest, model_ensemble)
+models <- list(model_rxlogit, model_rxforest,
+ model_rxtrees1, model_rxtrees2, model_rxtrees3, model_rxtrees4,
+ model_ensemble)
# Predict class
@@ -202,7 +260,7 @@ predictions <- lapply(models,
data=data_split[[2]]) %>%
lapply('[[', 1)
-levels(predictions[[4]]) <- c("no", "yes")
+levels(predictions[[7]]) <- c("no", "yes")
# Confusion matrix evaluation results.
@@ -237,7 +295,7 @@ pre_metrics <-
# Predict class probability
-probs <- lapply(models[c(1, 2, 3)],
+probs <- lapply(models[c(1, 2, 3, 4, 5, 6)],
rxPredict,
data=data_split[[2]]) %>%
lapply('[[', 3)
@@ -259,8 +317,18 @@ auc_metrics <- lapply(preds,
auc_metrics <- c(auc_metrics, NaN)
-algo_list <- c("rxLogisticRegression", "rxFastTrees", "rxFastForest", "rxEnsemble")
-time_consumption <- c(time_rxlogit[3], time_rxtrees[3], time_rxforest[[3]], time_ensemble[3])
+algo_list <- c("rxLogisticRegression",
+ "rxFastForest",
+ "rxFastTrees",
+ "rxFastTrees(500)",
+ "rxFastTrees(500, 0.3)",
+ "rxFastTrees(500, 0.3, ub)",
+ "rxEnsemble")
+
+time_consumption <- c(time_rxlogit[3], time_rxforest[[3]],
+ time_rxtrees1[3], time_rxtrees2[[3]],
+ time_rxtrees3[[3]], time_rxtrees4[[3]],
+ time_ensemble[3])
df_comp <-
data.frame(Models=algo_list,
@@ -269,6 +337,18 @@ df_comp <-
Precision=pre_metrics,
AUC=auc_metrics,
Time=time_consumption) %T>%
- {head(.) %>% print()}
+ print()
+```
+
+### 2.4 Save Models for Deployment
+
+Last but not least, we need to save the model objects in various formats, (e.g., `.RData`, `SQLServerData`, ect) for the later usage of deployment.
+
+```{r, message=FALSE, warning=FALSE, error=FALSE}
+# Save model for deployment usage.
+
+model_rxtrees <- model_rxtrees3
+
+save(model_rxtrees, file="model_rxtrees.RData")
```
diff --git a/CreditRiskPrediction/Code/CreditRiskScale.html b/CreditRiskPrediction/Code/CreditRiskScale.html
index 11d0c16..5d0277d 100644
--- a/CreditRiskPrediction/Code/CreditRiskScale.html
+++ b/CreditRiskPrediction/Code/CreditRiskScale.html
@@ -11,7 +11,7 @@
-
+
Faster and Scalable Credit Risk Prediction
@@ -119,7 +119,7 @@ $(document).ready(function () {
Faster and Scalable Credit Risk Prediction
Fang Zhou, Data Scientist, Microsoft
-2017-05-17
+2017-05-18
Now, let’s get started to build credit risk models by leveraging different machine learning algorithms from the MicrosoftML package.
First of all, we create individual machine learning models on the dataset processedSimu by using the functions rxLogisticRegression(), rxFastTrees(), and rxFastForest().
First of all, we create individual machine learning models on the dataset processedSimu.xdf by using the functions rxLogisticRegression(), rxFastForest(), rxFastTrees().
From the credit risk prediction template, we know that gradient boosting is the most suitable algorithm for this example, considering the overall performance. Therefore, the models implemented by the function rxFastTrees() with different sets of parameters are trained respectively.
## Variable roles.
# Target variable
@@ -228,9 +229,6 @@ vars <- setdiff(names(data), c(target, id))## Automatically adding a MinMax normalization transform, use 'norm=Warn' or 'norm=No' to turn this behavior off.
## LBFGS multi-threading will attempt to load dataset into memory. In case of out-of-memory issues, turn off multi-threading by setting trainThreads to 1.
@@ -271,30 +271,8 @@ time_rxlogit <- system.time(
## improvement criterion: Mean Improvement
## L1 regularization selected 13 of 17 weights.
## Not training a calibrator because it is not needed.
-## Elapsed time: 00:00:01.9451805
-## Elapsed time: 00:00:00.2341484# Train model: rxFastTrees
-
-time_rxtrees <- system.time(
-
- model_rxtrees <- rxFastTrees(
- formula=form,
- data=data_split[[1]],
- type="binary",
- numTrees=100,
- numLeaves=20,
- learningRate=0.2,
- minSplit=10)
-)## Not adding a normalizer.
-## Making per-feature arrays
-## Changing data from row-wise to column-wise
-## Processed 129155 instances
-## Binning and forming Feature objects
-## Reserved memory for tree learner: 219024 bytes
-## Starting to train ...
-## Not training a calibrator because it is not needed.
-## Elapsed time: 00:00:01.4975066# Train model: rxFastForest
time_rxforest <- system.time(
@@ -305,18 +283,73 @@ time_rxforest <- system.time(
type="binary",
numTrees=100,
numLeaves=20,
- minSplit=10)
+ minSplit=10,
+ verbose=0
+ )
+)
+
+# Train model: rxFastTrees
+
+time_rxtrees1 <- system.time(
+
+ model_rxtrees1 <- rxFastTrees(
+ formula=form,
+ data=data_split[[1]],
+ type="binary",
+ numTrees=100,
+ numLeaves=20,
+ learningRate=0.2,
+ minSplit=10,
+ unbalancedSets=FALSE,
+ verbose=0
+ )
+)
+
+time_rxtrees2 <- system.time(
+
+ model_rxtrees2 <- rxFastTrees(
+ formula=form,
+ data=data_split[[1]],
+ type="binary",
+ numTrees=500,
+ numLeaves=20,
+ learningRate=0.2,
+ minSplit=10,
+ unbalancedSets=FALSE,
+ verbose=0
+ )
+)
+
+time_rxtrees3 <- system.time(
+
+ model_rxtrees3 <- rxFastTrees(
+ formula=form,
+ data=data_split[[1]],
+ type="binary",
+ numTrees=500,
+ numLeaves=20,
+ learningRate=0.3,
+ minSplit=10,
+ unbalancedSets=FALSE,
+ verbose=0
+ )
+)
+
+time_rxtrees4 <- system.time(
+
+ model_rxtrees4 <- rxFastTrees(
+ formula=form,
+ data=data_split[[1]],
+ type="binary",
+ numTrees=500,
+ numLeaves=20,
+ learningRate=0.3,
+ minSplit=10,
+ unbalancedSets=TRUE,
+ verbose=0
+ )
)## Not adding a normalizer.
-## Making per-feature arrays
-## Changing data from row-wise to column-wise
-## Processed 129155 instances
-## Binning and forming Feature objects
-## Reserved memory for tree learner: 219024 bytes
-## Starting to train ...
-## Training calibrator.
-## Elapsed time: 00:00:02.2617353Next, we build an ensemble of models by using the function rxEnsemble().
Next, we build an ensemble of fast tree models by using the function rxEnsemble().
# Train an ensemble model.
time_ensemble <- system.time(
@@ -325,18 +358,25 @@ time_ensemble <- system.time(
formula=form,
data=data_split[[1]],
type="binary",
- trainers=list(logisticRegression(), fastTrees(), fastForest()),
- replace=TRUE
+ trainers=list(fastTrees(),
+ fastTrees(numTrees=500),
+ fastTrees(numTrees=500, learningRate=0.3),
+ fastTrees(numTrees=500, learningRate=0.3, unbalancedSets=TRUE)),
+ combineMethod="vote",
+ replace=TRUE,
+ verbose=0
)
)## Elapsed time: 00:00:00.4024905## Elapsed time: 00:00:00.6069022Finally, we evaluate and compare the above built models.
+Finally, we evaluate and compare the above built models at various aspects.
# Predict
-models <- list(model_rxlogit, model_rxtrees, model_rxforest, model_ensemble)
+models <- list(model_rxlogit, model_rxforest,
+ model_rxtrees1, model_rxtrees2, model_rxtrees3, model_rxtrees4,
+ model_ensemble)
# Predict class
@@ -344,11 +384,14 @@ predictions <- lapply(models,
rxPredict,
data=data_split[[2]]) %>%
lapply('[[', 1)## Elapsed time: 00:00:00.3329582
-## Elapsed time: 00:00:00.5682058
-## Elapsed time: 00:00:00.5726077
-## Elapsed time: 00:00:00.9745613levels(predictions[[4]]) <- c("no", "yes")
+## Elapsed time: 00:00:00.3323721
+## Elapsed time: 00:00:00.6021822
+## Elapsed time: 00:00:00.5783706
+## Elapsed time: 00:00:03.2957191
+## Elapsed time: 00:00:03.1592642
+## Elapsed time: 00:00:03.3192340
+## Elapsed time: 00:00:05.7721651
+levels(predictions[[7]]) <- c("no", "yes")
# Confusion matrix evaluation results.
@@ -383,13 +426,16 @@ pre_metrics <-
# Predict class probability
-probs <- lapply(models[c(1, 2, 3)],
+probs <- lapply(models[c(1, 2, 3, 4, 5, 6)],
rxPredict,
data=data_split[[2]]) %>%
lapply('[[', 3)
-## Elapsed time: 00:00:00.2420659
-## Elapsed time: 00:00:00.5683409
-## Elapsed time: 00:00:00.5884139
+## Elapsed time: 00:00:00.2500800
+## Elapsed time: 00:00:00.6540367
+## Elapsed time: 00:00:00.5695867
+## Elapsed time: 00:00:03.3178125
+## Elapsed time: 00:00:03.1851176
+## Elapsed time: 00:00:03.2898501
# Create prediction object
preds <- lapply(probs,
@@ -407,8 +453,18 @@ auc_metrics <- lapply(preds,
auc_metrics <- c(auc_metrics, NaN)
-algo_list <- c("rxLogisticRegression", "rxFastTrees", "rxFastForest", "rxEnsemble")
-time_consumption <- c(time_rxlogit[3], time_rxtrees[3], time_rxforest[[3]], time_ensemble[3])
+algo_list <- c("rxLogisticRegression",
+ "rxFastForest",
+ "rxFastTrees",
+ "rxFastTrees(500)",
+ "rxFastTrees(500, 0.3)",
+ "rxFastTrees(500, 0.3, ub)",
+ "rxEnsemble")
+
+time_consumption <- c(time_rxlogit[3], time_rxforest[[3]],
+ time_rxtrees1[3], time_rxtrees2[[3]],
+ time_rxtrees3[[3]], time_rxtrees4[[3]],
+ time_ensemble[3])
df_comp <-
data.frame(Models=algo_list,
@@ -417,12 +473,24 @@ df_comp <-
Precision=pre_metrics,
AUC=auc_metrics,
Time=time_consumption) %T>%
- {head(.) %>% print()}
-## Models Accuracy Recall Precision AUC Time
-## 1 rxLogisticRegression 0.9712880 0.0000000 NaN 0.5700583 2.36
-## 2 rxFastTrees 0.9807982 0.4589128 0.7823276 0.9383639 1.61
-## 3 rxFastForest 0.9712880 0.0000000 NaN 0.6735370 2.36
-## 4 rxEnsemble 0.9712880 0.0000000 NaN NaN 6.28
+ print()## Models Accuracy Recall Precision AUC Time
+## 1 rxLogisticRegression 0.9710578 0.0000000 NaN 0.5808235 2.27
+## 2 rxFastForest 0.9710578 0.0000000 NaN 0.6917648 2.57
+## 3 rxFastTrees 0.9805252 0.4527363 0.7827957 0.9392378 1.47
+## 4 rxFastTrees(500) 0.9877968 0.6915423 0.8593509 0.9833151 4.57
+## 5 rxFastTrees(500, 0.3) 0.9873108 0.7226368 0.8177340 0.9675594 4.42
+## 6 rxFastTrees(500, 0.3, ub) 0.9848989 0.9689055 0.6638262 0.9948528 4.23
+## 7 rxEnsemble 0.9833510 0.4471393 0.9523179 NaN 11.40Last but not least, we need to save the model objects in various formats, (e.g., .RData, SQLServerData, ect) for the later usage of deployment.
# Save model for deployment usage.
+
+model_rxtrees <- model_rxtrees3
+
+save(model_rxtrees, file="model_rxtrees.RData")