Introduction to Prediction Explanations

Peter Hurford, Colin Priest, Sergey Yurgenson, Thakur Raj Anand

2022-05-05

A few questions always asked by business leaders after seeing the results of highly accurate machine learning models are as follows
- Are machine learning models interpretable and transparent?
- How can the results of the model be used to develop a business strategy?
- Can the predictions from the model be used to explain to the regulators why something was rejected or accepted based on model prediction?

DataRobot does provide many diagnostics like partial dependence, feature impact, and prediction explanations to answer the above questions and using those diagnostics predictions can be converted to prescriptions for the business. In this vignette we would be covering prediction explanations. Partial dependence has been covered in detail in the companion vignette “Interpreting Predictive Models Using Partial Dependence Plots”.

Introduction

The DataRobot modeling engine is a commercial product that supports the rapid development and evaluation of a large number of different predictive models from a single data source. The open-source R package datarobot allows users of the DataRobot modeling engine to interact with it from R, creating new modeling projects, examining model characteristics, and generating predictions from any of these models for a specified dataset. This vignette illustrates how to interact with DataRobot using datarobot package, build models, make prediction using a model and then use prediction explanations to explain why a model is predicting high or low. Prediction explanations can be used to answer the questions mentioned earlier.

Load the useful libraries

Let’s load datarobot and other useful packages

library(httr)
library(knitr)
library(data.table)

Connecting to DataRobot

To access the DataRobot modeling engine, it is necessary to establish an authenticated connection, which can be done in one of two ways. In both cases, the necessary information is an endpoint - the URL address of the specific DataRobot server being used - and a token, a previously validated access token.

token is unique for each DataRobot modeling engine account and can be accessed using the DataRobot webapp in the account profile section.

endpoint depends on DataRobot modeling engine installation (cloud-based, on-prem…) you are using. Contact your DataRobot admin for endpoint to use. The endpoint for DataRobot cloud accounts is https://app.datarobot.com/api/v2

The first access method uses a YAML configuration file with these two elements - labeled token and endpoint - located at $HOME/.config/datarobot/drconfig.yaml. If this file exists when the datarobot package is loaded, a connection to the DataRobot modeling engine is automatically established. It is also possible to establish a connection using this YAML file via the ConnectToDataRobot function, by specifying the configPath parameter.

The second method of establishing a connection to the DataRobot modeling engine is to call the function ConnectToDataRobot with the endpoint and token parameters.

library(datarobot)
endpoint <- "https://<YOUR ENDPOINT HERE>/api/v2"
apiToken <- "<YOUR API TOKEN HERE>"
ConnectToDataRobot(endpoint = endpoint, token = apiToken)

Data

We would be using a sample dataset related to credit scoring open sourced by LendingClub (https://www.lendingclub.com/). Below is the information related to the variables.

loan_amnt Min. : 500 1st Qu.: 5000 Median : 9600 Mean :11036 3rd Qu.:15000 Max. :35000 NA
funded_amnt Min. : 500 1st Qu.: 5000 Median : 9250 Mean :10766 3rd Qu.:15000 Max. :35000 NA
term Length:10000 Class :character Mode :character NA NA NA NA
int_rate Length:10000 Class :character Mode :character NA NA NA NA
installment Min. : 15.69 1st Qu.: 163.47 Median : 275.11 Mean : 321.43 3rd Qu.: 427.04 Max. :1276.60 NA
grade Length:10000 Class :character Mode :character NA NA NA NA
sub_grade Length:10000 Class :character Mode :character NA NA NA NA
emp_title Length:10000 Class :character Mode :character NA NA NA NA
emp_length Length:10000 Class :character Mode :character NA NA NA NA
home_ownership Length:10000 Class :character Mode :character NA NA NA NA
annual_inc Min. : 2000 1st Qu.: 40000 Median : 58000 Mean : 68203 3rd Qu.: 82000 Max. :900000 NA’s :1
verification_status Length:10000 Class :character Mode :character NA NA NA NA
pymnt_plan Length:10000 Class :character Mode :character NA NA NA NA
url Length:10000 Class :character Mode :character NA NA NA NA
desc Length:10000 Class :character Mode :character NA NA NA NA
purpose Length:10000 Class :character Mode :character NA NA NA NA
title Length:10000 Class :character Mode :character NA NA NA NA
zip_code Length:10000 Class :character Mode :character NA NA NA NA
addr_state Length:10000 Class :character Mode :character NA NA NA NA
dti Min. : 0.00 1st Qu.: 8.16 Median :13.41 Mean :13.34 3rd Qu.:18.69 Max. :29.99 NA
delinq_2yrs Min. : 0.0000 1st Qu.: 0.0000 Median : 0.0000 Mean : 0.1482 3rd Qu.: 0.0000 Max. :11.0000 NA’s :5
earliest_cr_line Length:10000 Class :character Mode :character NA NA NA NA
inq_last_6mths Min. : 0.000 1st Qu.: 0.000 Median : 1.000 Mean : 1.067 3rd Qu.: 2.000 Max. :25.000 NA’s :5
mths_since_last_delinq Min. : 0.00 1st Qu.: 18.00 Median : 34.00 Mean : 35.89 3rd Qu.: 53.00 Max. :120.00 NA’s :6316
mths_since_last_record Min. : 0.00 1st Qu.: 0.00 Median : 86.00 Mean : 61.65 3rd Qu.:101.00 Max. :119.00 NA’s :9160
open_acc Min. : 1.000 1st Qu.: 6.000 Median : 9.000 Mean : 9.335 3rd Qu.:12.000 Max. :39.000 NA’s :5
pub_rec Min. :0.00000 1st Qu.:0.00000 Median :0.00000 Mean :0.06013 3rd Qu.:0.00000 Max. :3.00000 NA’s :5
revol_bal Min. : 0 1st Qu.: 3524 Median : 8646 Mean : 14271 3rd Qu.: 16952 Max. :1207359 NA
revol_util Min. : 0.00 1st Qu.: 25.00 Median : 48.70 Mean : 48.47 3rd Qu.: 71.90 Max. :108.80 NA’s :23
total_acc Min. : 1.00 1st Qu.:13.00 Median :20.00 Mean :22.09 3rd Qu.:29.00 Max. :90.00 NA’s :5
initial_list_status Length:10000 Class :character Mode :character NA NA NA NA
mths_since_last_major_derog Mode:logical NA’s:10000 NA NA NA NA NA
policy_code Min. :1 1st Qu.:1 Median :1 Mean :1 3rd Qu.:1 Max. :1 NA
is_bad Min. :0.0000 1st Qu.:0.0000 Median :0.0000 Mean :0.1295 3rd Qu.:0.0000 Max. :1.0000 NA

Divide data into train and test and setup the project

Let’s divide our data in train and test. We can use train data to create a datarobot project using StartProject function and test data to make predictions and generate prediction explanations. Detailed explanation about creating projects was described in the vignette , “Introduction to the DataRobot R Package.” The specific sequence used here was:

target <- "is_bad"
projectName <- "Credit Scoring"

set.seed(1111)
split <- sample(nrow(Lending), round(0.9 * nrow(Lending)), replace = FALSE)
train <- Lending[split,]
test <- Lending[-split,]

project <- StartProject(dataSource = train, 
                        projectName = projectName,
                        target = target,
                        workerCount = "max",
                        wait = TRUE)

Once the modeling process has completed, the ListModels function returns an S3 object of class “listOfModels” that characterizes all of the models in a specified DataRobot project.

results <- as.data.frame(ListModels(project))
kable(head(results), longtable = TRUE, booktabs = TRUE, row.names = TRUE)
modelType expandedModel modelId blueprintId featurelistName featurelistId samplePct validationMetric
1 ENET Blender ENET Blender 5cdaff5621538736b976979e d451f445e32c473bd7250b175e2a8759 Informative Features 5cdafa9c319c7b34caeeeb75 64 0.32417
2 Gradient Boosted Greedy Trees Classifier with Early Stopping Gradient Boosted Greedy Trees Classifier with Early Stopping::One-Hot Encoding::Univariate credibility estimates with ElasticNet::Category Count::Converter for Text Mining::Auto-Tuned Word N-Gram Text Modeler using token occurrences::Missing Values Imputed::Search for differences::Search for ratios 5cdafc69215387127c7697ac 788bd071d49fc43acc55e646c3341445 Informative Features 5cdafa9c319c7b34caeeeb75 64 0.32456
3 ENET Blender ENET Blender 5cdaff5621538736b97697a0 64080ee61be8e47ab6e09c35d9868972 Informative Features 5cdafa9c319c7b34caeeeb75 64 0.32475
4 Advanced AVG Blender Advanced AVG Blender 5cdaff5521538736b976979c 00790837050987b652403c758b2a90f1 Informative Features 5cdafa9c319c7b34caeeeb75 64 0.32485
5 Gradient Boosted Greedy Trees Classifier with Early Stopping Gradient Boosted Greedy Trees Classifier with Early Stopping::One-Hot Encoding::Univariate credibility estimates with ElasticNet::Category Count::Converter for Text Mining::Auto-Tuned Word N-Gram Text Modeler using token occurrences::Missing Values Imputed::Search for differences::Search for ratios 5cdafe722153872da07697be 788bd071d49fc43acc55e646c3341445 Informative Features 5cdafa9c319c7b34caeeeb75 80 0.32549
6 AVG Blender AVG Blender 5cdaff5521538736b976979a fe81e57d50d41f91322af82b7973d878 Informative Features 5cdafa9c319c7b34caeeeb75 64 0.32558

Generating Model Predictions

Let’s look at some model predictions. The generation of model predictions uses the Predict function:

bestModel <- GetRecommendedModel(project)
bestPredictions <- Predict(bestModel, test, type = "probability")
testPredictions <- data.frame(original = test$is_bad, prediction = bestPredictions)
kable(head(testPredictions), longtable = TRUE, booktabs = TRUE, row.names = TRUE)
original prediction
1 0 0.0714697
2 0 0.1396974
3 0 0.0925635
4 0 0.0593510
5 0 0.1172840
6 0 0.1049122

Calculate Prediction Explanations

For each prediction, DataRobot provides an ordered list of explanations; the number of explanations is based on the setting. Each explanation is a feature from the dataset and its corresponding value, accompanied by a qualitative indicator of the explanation’s strength—strong (+++), medium (++), or weak (+) positive or negative (-) influence.

There are three main inputs you can set for DataRobot to use when computing prediction explanations
1. maxExplanations: the Number of explanations for each predictions. Default is 3.
2. thresholdLow: Probability threshold below which DataRobot should calculate prediction explanations.
3. thresholdHigh: Probability threshold above which DataRobot should calculate prediction explanations.

explanations <- GetPredictionExplanations(bestModel, test, maxExplanations = 3,
                                          thresholdLow = 0.25, thresholdHigh = 0.75)
kable(head(explanations), longtable = TRUE, booktabs = TRUE, row.names = TRUE)
rowId prediction class1Label class1Probability class2Label class2Probability reason1FeatureName reason1FeatureValue reason1QualitativeStrength reason1Strength reason1Label reason2FeatureName reason2FeatureValue reason2QualitativeStrength reason2Strength reason2Label reason3FeatureName reason3FeatureValue reason3QualitativeStrength reason3Strength reason3Label
1 0 0 1 0.0714697 0 0.9285303 revol_util 37.1 -0.2260173 1 Notes ++ 0.1487372 1 total_acc 23 -0.1403030 1
2 1 0 1 0.1396974 0 0.8603026 inq_last_6mths 2 +++ 0.2362790 1 total_acc 27 -0.2090679 1 purpose Consolidation ++ 0.1447419 1
3 2 0 1 0.0925635 0 0.9074365 annual_inc 40000 +++ 0.3073045 1 purpose FICO score 762 want’s to buy a new car -0.2362631 1 revol_util 18.3 -0.2127375 1
4 3 0 1 0.0593510 0 0.9406490 revol_util 39.5 -0.2576392 1 total_acc 24 -0.2124569 1 annual_inc 112000 -0.2024626 1
5 4 0 1 0.1172840 0 0.8827160 revol_util 62.1 +++ 0.2626324 1 earliest_cr_line (Day of Week) 2 -0.1870957 1 total_acc 19 -0.1591037 1
6 5 0 1 0.1049122 0 0.8950878 inq_last_6mths 3 ++ 0.2975595 1 total_acc 24 -0.2587315 1 purpose Paying off a personal loan ++ 0.2087095 1

From the example above, you could answer “Why did the model give one of the customers a 97% probability of defaulting?” Top explanation explains that purpose_cat of loan was “credit card small business”” and we can also see in above example that whenever model is predicting high probability of default, purpose_cat is related to small business.

Some notes on explanations:
- If the data points are very similar, the explanations can list the same rounded up values.
- It is possible to have a explanation state of MISSING if a “missing value” was important in making the prediction.
- Typically, the top explanations for a prediction have the same direction as the outcome, but it’s possible that with interaction effects or correlations among variables a explanation could, for instance, have a strong positive impact on a negative prediction.

Adjusted Predictions in Prediction Explanations

In some projects – such as insurance projects – the prediction adjusted by exposure is more useful to look at than just raw prediction. For example, the raw prediction (e.g. claim counts) is divided by exposure (e.g. time) in the project with exposure column. The adjusted prediction provides insights with regard to the predicted claim counts per unit of time. To include that information, set excludeAdjustedPredictions to False in correspondent method calls.

explanations <- GetPredictionExplanations(bestModel, test, maxExplanations = 3,
                                          thresholdLow = 0.25, thresholdHigh = 0.75,
                                          excludeAdjustedPredictions = FALSE)
kable(head(explanations), longtable = TRUE, booktabs = TRUE, row.names = TRUE)
rowId prediction class1Label class1Probability class2Label class2Probability reason1FeatureName reason1FeatureValue reason1QualitativeStrength reason1Strength reason1Label reason2FeatureName reason2FeatureValue reason2QualitativeStrength reason2Strength reason2Label reason3FeatureName reason3FeatureValue reason3QualitativeStrength reason3Strength reason3Label
1 0 0 1 0.0714697 0 0.9285303 revol_util 37.1 -0.2260173 1 Notes ++ 0.1487372 1 total_acc 23 -0.1403030 1
2 1 0 1 0.1396974 0 0.8603026 inq_last_6mths 2 +++ 0.2362790 1 total_acc 27 -0.2090679 1 purpose Consolidation ++ 0.1447419 1
3 2 0 1 0.0925635 0 0.9074365 annual_inc 40000 +++ 0.3073045 1 purpose FICO score 762 want’s to buy a new car -0.2362631 1 revol_util 18.3 -0.2127375 1
4 3 0 1 0.0593510 0 0.9406490 revol_util 39.5 -0.2576392 1 total_acc 24 -0.2124569 1 annual_inc 112000 -0.2024626 1
5 4 0 1 0.1172840 0 0.8827160 revol_util 62.1 +++ 0.2626324 1 earliest_cr_line (Day of Week) 2 -0.1870957 1 total_acc 19 -0.1591037 1
6 5 0 1 0.1049122 0 0.8950878 inq_last_6mths 3 ++ 0.2975595 1 total_acc 24 -0.2587315 1 purpose Paying off a personal loan ++ 0.2087095 1

Summary

This note has described the Prediction Explanations which are useful for understanding why model is predicting high or low predictions for a specific case. DataRobot also provides qualitative strength of each explanation. Prediction Explanations can be used in developing good business strategy by taking prescriptions based on the explanations which are responsible for high or low predictions. They are also useful in explaining the actions taken based on the model predictions to regulatory or compliance department within an organization.