README.md
In ben519/claimz: Fast utility methods for insurance claims data using data.table
claimz
Fast utility methods for insurance claims data using data.table
About
The goals of this package are
- Provide a framework for wrangling and analyzing claims data
- Provide methods for validating structure (e.g. every claim should be tied to a policy)
Installation
You can install claimz from github with:
# install.packages("devtools")
devtools::install_github("ben519/claimz")
Demonstration
Load packages
library(claimz)
library(data.table)
Toy datasets included in the claimz package
policiez
PolicyID EffectiveDate ExpirationDate
1: 1 2014-01-01 2015-01-01
2: 2 2014-06-01 2015-06-01
claimz
ClaimID PolicyID EffectiveDate ExpirationDate DateOfLoss ReportDate
1: 1 1 2014-01-01 2015-01-01 2014-03-15 2014-03-15
2: 2 1 2014-01-01 2015-01-01 2014-04-01 2015-07-11
3: 3 2 2014-06-01 2015-06-01 2015-05-30 2015-06-10
claimvaluationz
ValuationDate ClaimID PolicyID EffectiveDate ExpirationDate DateOfLoss ReportDate Incurred Paid
1: 2015-01-01 1 1 2014-01-01 2015-01-01 2014-03-15 2014-03-15 100 50
2: 2015-01-01 2 1 2014-01-01 2015-01-01 2014-04-01 2015-07-11 150 100
3: 2016-01-01 1 1 2014-01-01 2015-01-01 2014-03-15 2014-03-15 125 125
4: 2016-01-01 2 1 2014-01-01 2015-01-01 2014-04-01 2015-07-11 230 230
5: 2016-01-01 3 2 2014-06-01 2015-06-01 2015-05-30 2015-06-10 75 75
Validate the data structure and relationships
is_valid_datasets(policiez, claimz, claimvaluationz) # No warnings
is_valid_datasets(policiez, head(claimz, 2), claimvaluationz) # 1 warning: "1 unique ClaimIDs in claimvaluations not in claims"
Claim Snapshots
We can run special queries of claim-valuations using the function claims_at(...). For example, we can get all claims valued as of 2015-06-30.
claims_at(claimvaluationz, valuationDate = as.Date("2015-06-30"))
ClaimID ValuationDate CHValuationDate
1: 1 2015-06-30 2015-01-01
2: 2 2015-06-30 2015-01-01
3: 3 2015-06-30 <NA>
Take a look at ClaimID = 1. It has date of loss (DOL) = 2014-03-15, and subsequent claim valuation records on 2015-01-01 and 2016-01-01. For this claim, claims_at(...) assumes that the value of the claim on 2015-06-30 was the same as its value on 2015-01-01 since this is the closest valuation date for Claim 1 prior to the desired valuation date. The pair of columns (ClaimID, CHValuationDate) can be used to map each row in the result to the corresponding row from claimvaluations.
Now take a look at claim 3. It has DOL = 2015-05-30, and one valuation record on 2016-01-01. So, as of 2015-06-30 the claim had occured, but the insurance company ws unaware of existence. This is why it is returned with CHValuationDate = NA.
We can also query claims as of a particular age. For example we can get all claims as of age = 10 months.
claims_at(claimvaluationz, claimAge = 10)
ClaimID ValuationDate CHValuationDate DateOfLoss
1: 1 2015-01-15 2015-01-01 2014-03-15
2: 2 2015-02-01 2015-01-01 2014-04-01
3: 3 2016-03-30 2016-01-01 2015-05-30
Here, age is measured in months since the DOL of the claim. For example, ClaimID = 1 occured on 2014-03-15. The claim is 10 months old as of 2015-01-15, so the value of the claim as of this date is returned. Of course, claimvaluations does not contain a row for (ClaimID = 1, ValuationDate = 2015-01-15) so the most recent prior valuation point for the claim is assumed to be the value as of that date.
Sometimes this can be problematic. For example, consider ClaimID = 3. It is 10 months old as of 2016-03-30, but perhaps our data is only valid up through 2016-02-29. In this case, resulting row for (ClaimID = 3, ValuationDate = 2016-03-30) is invalid because it is immature. We can prevent this from happening by using the parameters maxValuationDate = as.Date("2016-02-29") and dropImmatureValuations = TRUE (which is TRUE by default).
claims_at(claimvaluationz, claimAge = 10, maxValuationDate = as.Date("2016-02-29"), dropImmatureValuations = TRUE)
ClaimID ValuationDate CHValuationDate DateOfLoss
1: 1 2015-01-15 2015-01-01 2014-03-15
2: 2 2015-02-01 2015-01-01 2014-04-01
(Note that Claim 3 is exluded from this result set).
We might also want to get the value of each claim when its corresponding policy is a certain age. For example
claims_at(claimvaluationz, policyAge = 18)
ClaimID ValuationDate CHValuationDate EffectiveDate
1: 1 2015-07-01 2015-01-01 2014-01-01
2: 2 2015-07-01 2015-01-01 2014-01-01
3: 3 2015-12-01 <NA> 2014-06-01
Looking at ClaimID = 2, its policy began on 2014-01-01, so the policy was 18 months old as of 2015-07-01. The result shows that it can be mapped back to the 2015-01-01 valuation point in claimvaluationz to infer the values of the claim when the policy was 18 months old.
Loss Triangles
The claimz package also makes it easy to generate loss triangles for analyzing changes in loss amounts over time. To start, we'll make use of the simulate_losses() method to generate a random set of claimvaluations.
set.seed(4)
claimvalz <- simulate_losses(numLosses = 10, startDate = as.Date("2010-01-1"), endDate = as.Date("2012-12-31"))
claimvalz
| ClaimID | DateOfLoss | ReportDate | CloseDate | ValuationDate | Paid |
|:-------:|:----------:|:----------:|:----------:|:-------------:|:--------:|
| 1 | 2010-01-11 | 2010-01-11 | NA | 2010-01-18 | 857.71 |
| 1 | 2010-01-11 | 2010-01-11 | NA | 2010-07-09 | 967.22 |
| 1 | 2010-01-11 | 2010-01-11 | NA | 2010-08-07 | 1566.26 |
| 1 | 2010-01-11 | 2010-01-11 | NA | 2010-10-22 | 4276.51 |
| 1 | 2010-01-11 | 2010-01-11 | NA | 2011-05-07 | 5097.13 |
| ... | ... | ... | ... | ... | ... |
| 7 | 2012-03-05 | 2012-03-05 | NA | 2012-12-31 | 783.25 |
| 8 | 2012-09-21 | 2012-09-21 | NA | 2012-10-13 | 5291.33 |
| 8 | 2012-09-21 | 2012-09-21 | NA | 2012-10-21 | 6083.48 |
| 8 | 2012-09-21 | 2012-09-21 | NA | 2012-12-04 | 6114.92 |
| 9 | 2012-11-07 | 2012-11-07 | 2012-11-07 | 2012-11-07 | 29385.51 |
The function make_triangles() has a host of parameters for customizing loss triangles. For example, if we wanted to generate annual triangles, we could simply run make_triangles(claimvalz, minLeftOrigin = as.Date("2010-01-01"), originLength = 12). The result is a list of six triangles.
make_triangles(claimvalz, minLeftOrigin = as.Date("2010-01-01"), originLength = 12)
$Occurred.cmltv
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 5 5 5
2011-01-01 - 2011-12-31 1 1 NA
2012-01-01 - 2012-12-31 3 NA NA
$Occurred
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 5 0 0
2011-01-01 - 2011-12-31 1 0 NA
2012-01-01 - 2012-12-31 3 NA NA
$Reported.cmltv
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 4 5 5
2011-01-01 - 2011-12-31 1 1 NA
2012-01-01 - 2012-12-31 3 NA NA
$Reported
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 4 1 0
2011-01-01 - 2011-12-31 1 0 NA
2012-01-01 - 2012-12-31 3 NA NA
$Paid
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 41751.49 49080.76 49080.76
2011-01-01 - 2011-12-31 6189.03 6189.03 NA
2012-01-01 - 2012-12-31 36283.68 NA NA
$Paid.chg
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 41751.49 7329.27 0
2011-01-01 - 2011-12-31 6189.03 0.00 NA
2012-01-01 - 2012-12-31 36283.68 NA NA
Each row represents a distinct set of claims. For example, the first row, 2010-01-01 - 2010-12-31, represents claims that occurred between 2010-01-01 and 2010-12-31. Looking at the Reported triangle, we see that four claims which occurred in the period were reported within 12 months of 2010-01-01 and 1 other claim the occurred in the period was reported between 12 and 24 months. Similarly, row two shows that one claim which occurred between 2011-01-01 and 2011-12-31 was reported within 12 months of 2011-01-01.
We could also look at quarterly or monthly triangles
# quarterly
make_triangles(claimvalz, minLeftOrigin = as.Date("2010-01-01"), originLength = 3, rowDev = 3, colDev = 3)
# monthly
make_triangles(claimvalz, minLeftOrigin = as.Date("2010-01-01"), originLength = 1, rowDev = 1, colDev = 1)
There are many additional options within make_triangles() - too many to explain here so be sure to read the function documentation to learn all its capabilities.
Contact
If you'd like to contact me regarding bugs, questions, or general consulting, feel free to drop me a line - bgorman519@gmail.com
ben519/claimz documentation built on May 12, 2019, 10:56 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
claimz
Fast utility methods for insurance claims data using data.table
About
The goals of this package are
- Provide a framework for wrangling and analyzing claims data
- Provide methods for validating structure (e.g. every claim should be tied to a policy)
Installation
You can install claimz from github with:
# install.packages("devtools")
devtools::install_github("ben519/claimz")
Demonstration
Load packages
library(claimz)
library(data.table)
Toy datasets included in the claimz package
policiez
PolicyID EffectiveDate ExpirationDate
1: 1 2014-01-01 2015-01-01
2: 2 2014-06-01 2015-06-01
claimz
ClaimID PolicyID EffectiveDate ExpirationDate DateOfLoss ReportDate
1: 1 1 2014-01-01 2015-01-01 2014-03-15 2014-03-15
2: 2 1 2014-01-01 2015-01-01 2014-04-01 2015-07-11
3: 3 2 2014-06-01 2015-06-01 2015-05-30 2015-06-10
claimvaluationz
ValuationDate ClaimID PolicyID EffectiveDate ExpirationDate DateOfLoss ReportDate Incurred Paid
1: 2015-01-01 1 1 2014-01-01 2015-01-01 2014-03-15 2014-03-15 100 50
2: 2015-01-01 2 1 2014-01-01 2015-01-01 2014-04-01 2015-07-11 150 100
3: 2016-01-01 1 1 2014-01-01 2015-01-01 2014-03-15 2014-03-15 125 125
4: 2016-01-01 2 1 2014-01-01 2015-01-01 2014-04-01 2015-07-11 230 230
5: 2016-01-01 3 2 2014-06-01 2015-06-01 2015-05-30 2015-06-10 75 75
Validate the data structure and relationships
is_valid_datasets(policiez, claimz, claimvaluationz) # No warnings
is_valid_datasets(policiez, head(claimz, 2), claimvaluationz) # 1 warning: "1 unique ClaimIDs in claimvaluations not in claims"
Claim Snapshots
We can run special queries of claim-valuations using the function claims_at(...). For example, we can get all claims valued as of 2015-06-30.
claims_at(claimvaluationz, valuationDate = as.Date("2015-06-30"))
ClaimID ValuationDate CHValuationDate
1: 1 2015-06-30 2015-01-01
2: 2 2015-06-30 2015-01-01
3: 3 2015-06-30 <NA>
Take a look at ClaimID = 1. It has date of loss (DOL) = 2014-03-15, and subsequent claim valuation records on 2015-01-01 and 2016-01-01. For this claim, claims_at(...) assumes that the value of the claim on 2015-06-30 was the same as its value on 2015-01-01 since this is the closest valuation date for Claim 1 prior to the desired valuation date. The pair of columns (ClaimID, CHValuationDate) can be used to map each row in the result to the corresponding row from claimvaluations.
Now take a look at claim 3. It has DOL = 2015-05-30, and one valuation record on 2016-01-01. So, as of 2015-06-30 the claim had occured, but the insurance company ws unaware of existence. This is why it is returned with CHValuationDate = NA.
We can also query claims as of a particular age. For example we can get all claims as of age = 10 months.
claims_at(claimvaluationz, claimAge = 10)
ClaimID ValuationDate CHValuationDate DateOfLoss
1: 1 2015-01-15 2015-01-01 2014-03-15
2: 2 2015-02-01 2015-01-01 2014-04-01
3: 3 2016-03-30 2016-01-01 2015-05-30
Here, age is measured in months since the DOL of the claim. For example, ClaimID = 1 occured on 2014-03-15. The claim is 10 months old as of 2015-01-15, so the value of the claim as of this date is returned. Of course, claimvaluations does not contain a row for (ClaimID = 1, ValuationDate = 2015-01-15) so the most recent prior valuation point for the claim is assumed to be the value as of that date.
Sometimes this can be problematic. For example, consider ClaimID = 3. It is 10 months old as of 2016-03-30, but perhaps our data is only valid up through 2016-02-29. In this case, resulting row for (ClaimID = 3, ValuationDate = 2016-03-30) is invalid because it is immature. We can prevent this from happening by using the parameters maxValuationDate = as.Date("2016-02-29") and dropImmatureValuations = TRUE (which is TRUE by default).
claims_at(claimvaluationz, claimAge = 10, maxValuationDate = as.Date("2016-02-29"), dropImmatureValuations = TRUE)
ClaimID ValuationDate CHValuationDate DateOfLoss
1: 1 2015-01-15 2015-01-01 2014-03-15
2: 2 2015-02-01 2015-01-01 2014-04-01
(Note that Claim 3 is exluded from this result set).
We might also want to get the value of each claim when its corresponding policy is a certain age. For example
claims_at(claimvaluationz, policyAge = 18)
ClaimID ValuationDate CHValuationDate EffectiveDate
1: 1 2015-07-01 2015-01-01 2014-01-01
2: 2 2015-07-01 2015-01-01 2014-01-01
3: 3 2015-12-01 <NA> 2014-06-01
Looking at ClaimID = 2, its policy began on 2014-01-01, so the policy was 18 months old as of 2015-07-01. The result shows that it can be mapped back to the 2015-01-01 valuation point in claimvaluationz to infer the values of the claim when the policy was 18 months old.
Loss Triangles
The claimz package also makes it easy to generate loss triangles for analyzing changes in loss amounts over time. To start, we'll make use of the simulate_losses() method to generate a random set of claimvaluations.
set.seed(4)
claimvalz <- simulate_losses(numLosses = 10, startDate = as.Date("2010-01-1"), endDate = as.Date("2012-12-31"))
claimvalz
| ClaimID | DateOfLoss | ReportDate | CloseDate | ValuationDate | Paid |
|:-------:|:----------:|:----------:|:----------:|:-------------:|:--------:|
| 1 | 2010-01-11 | 2010-01-11 | NA | 2010-01-18 | 857.71 |
| 1 | 2010-01-11 | 2010-01-11 | NA | 2010-07-09 | 967.22 |
| 1 | 2010-01-11 | 2010-01-11 | NA | 2010-08-07 | 1566.26 |
| 1 | 2010-01-11 | 2010-01-11 | NA | 2010-10-22 | 4276.51 |
| 1 | 2010-01-11 | 2010-01-11 | NA | 2011-05-07 | 5097.13 |
| ... | ... | ... | ... | ... | ... |
| 7 | 2012-03-05 | 2012-03-05 | NA | 2012-12-31 | 783.25 |
| 8 | 2012-09-21 | 2012-09-21 | NA | 2012-10-13 | 5291.33 |
| 8 | 2012-09-21 | 2012-09-21 | NA | 2012-10-21 | 6083.48 |
| 8 | 2012-09-21 | 2012-09-21 | NA | 2012-12-04 | 6114.92 |
| 9 | 2012-11-07 | 2012-11-07 | 2012-11-07 | 2012-11-07 | 29385.51 |
The function make_triangles() has a host of parameters for customizing loss triangles. For example, if we wanted to generate annual triangles, we could simply run make_triangles(claimvalz, minLeftOrigin = as.Date("2010-01-01"), originLength = 12). The result is a list of six triangles.
make_triangles(claimvalz, minLeftOrigin = as.Date("2010-01-01"), originLength = 12)
$Occurred.cmltv
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 5 5 5
2011-01-01 - 2011-12-31 1 1 NA
2012-01-01 - 2012-12-31 3 NA NA
$Occurred
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 5 0 0
2011-01-01 - 2011-12-31 1 0 NA
2012-01-01 - 2012-12-31 3 NA NA
$Reported.cmltv
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 4 5 5
2011-01-01 - 2011-12-31 1 1 NA
2012-01-01 - 2012-12-31 3 NA NA
$Reported
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 4 1 0
2011-01-01 - 2011-12-31 1 0 NA
2012-01-01 - 2012-12-31 3 NA NA
$Paid
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 41751.49 49080.76 49080.76
2011-01-01 - 2011-12-31 6189.03 6189.03 NA
2012-01-01 - 2012-12-31 36283.68 NA NA
$Paid.chg
Age
Origin 12 24 36
2010-01-01 - 2010-12-31 41751.49 7329.27 0
2011-01-01 - 2011-12-31 6189.03 0.00 NA
2012-01-01 - 2012-12-31 36283.68 NA NA
Each row represents a distinct set of claims. For example, the first row, 2010-01-01 - 2010-12-31, represents claims that occurred between 2010-01-01 and 2010-12-31. Looking at the Reported triangle, we see that four claims which occurred in the period were reported within 12 months of 2010-01-01 and 1 other claim the occurred in the period was reported between 12 and 24 months. Similarly, row two shows that one claim which occurred between 2011-01-01 and 2011-12-31 was reported within 12 months of 2011-01-01.
We could also look at quarterly or monthly triangles
# quarterly
make_triangles(claimvalz, minLeftOrigin = as.Date("2010-01-01"), originLength = 3, rowDev = 3, colDev = 3)
# monthly
make_triangles(claimvalz, minLeftOrigin = as.Date("2010-01-01"), originLength = 1, rowDev = 1, colDev = 1)
There are many additional options within make_triangles() - too many to explain here so be sure to read the function documentation to learn all its capabilities.
Contact
If you'd like to contact me regarding bugs, questions, or general consulting, feel free to drop me a line - bgorman519@gmail.com
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.