Type: Package
Title: Revenue Management Modeling
Version: 0.1.0
Date: 2022-04-29
Maintainer: Chul Kim <statkim7578@yonsei.ac.kr>
Description: The RMM fits Revenue Management Models using the RDE(Robust Demand Estimation) method introduced in the paper by <doi:10.2139/ssrn.3598259>, one of the customer choice-based Revenue Management Model. Furthermore, it is possible to select a multinomial model as well as a conditional logit model as a model of RDE.
License: GPL-2 | GPL-3 [expanded from: GPL (≥ 2.0)]
Encoding: UTF-8
LazyData: true
Depends: R (≥ 3.1.0)
Imports: tibble, magrittr, rlang, dplyr, numDeriv, Rcpp
RoxygenNote: 7.1.2
LinkingTo: Rcpp
NeedsCompilation: yes
Packaged: 2022-05-06 12:03:59 UTC; chulkim
Author: Chul Kim [aut, cre], Sanghoon Cho [aut], Jongho Im [aut]
Repository: CRAN
Date/Publication: 2022-05-09 11:30:05 UTC

Estimating total arrivals and the number of no purchase.

Description

Estimating total arrivals and the number of no purchase. Internal function that users should not call directly.

Usage

Arrival(
  CCB,
  Attr,
  Choice_Set_List,
  Alts_Same_Choice_Set,
  Is_Purchase,
  Base_Prod,
  Num_Prod,
  Num_Attr
)

Arguments

CCB

numeric vector, estimated model parameters(gamma and eta).

Attr

list, a set of data frame in which alternative specific variables.

Choice_Set_List

list, product code included for each choice set.

Alts_Same_Choice_Set

list, for internal calculation, a list of data exposed to the same choice set.

Is_Purchase

list, for internal calculation, list of customers who are exposed to the same choice set who purchased the product. If purchased 1, otherwise 0.

Base_Prod

numeric, the baseline product number

Num_Prod

numeric, the number of all products in observed data.

Num_Attr

numeric, the number of attributes in observed data.

Value

The estimated value of total arrivals and the number of no purchase.

Collects a choice set exposed to individuals.

Description

Collects a choice set exposed to individuals. Internal function that users should not call directly.

Usage

Choice_Set(df, num_id, uniq_id, idvar)

Arguments

df

a long format tibble.

num_id

Number of unique Booking_IDs appearing in transaction data

uniq_id

Unique Booking_ID in transaction data.

idvar

Variable name representing customer id (Booking_ID).

Value

Returns a list containing the values required for calculation within the rmm_reshape function.

Estimating model parameters.

Description

Estimating model parameters except gamma(no-purchase utility). Internal function that users should not call directly.

Usage

ESTF(Alts_Same_Choice_Set, Choice_Set_List, Num_Prod, Num_Attr, Is_Purchase)

Arguments

Alts_Same_Choice_Set

list, for internal calculation, a list of data exposed to the same choice set.

Choice_Set_List

list, product code included for each choice set.

Num_Prod

numeric, the number of all products in observed data.

Num_Attr

numeric, the number of attributes in observed data.

Is_Purchase

list, for internal calculation, list of customers who are exposed to the same choice set who purchased the product. If purchased 1, otherwise 0.

Value

The estimated value of parameters except gamma(no-purchase utility).

Estimating model parameters.

Description

Estimating model parameters except gamma(no-purchase utility) when a particular baseline_product is specified. Internal function that users should not call directly.

Usage

ESTF_bp(
  Base_Prod = 1,
  Choice_Set_List,
  Alts_Same_Choice_Set,
  Num_Prod,
  Num_Attr,
  Is_Purchase
)

Arguments

Base_Prod

numeric, the baseline product number

Choice_Set_List

list, product code included for each choice set.

Alts_Same_Choice_Set

list, for internal calculation, a list of data exposed to the same choice set.

Num_Prod

numeric, the number of all products in observed data.

Num_Attr

numeric, the number of attributes in observed data.

Is_Purchase

list, for internal calculation, list of customers who are exposed to the same choice set who purchased the product. If purchased 1, otherwise 0.

Value

The estimated value of parameters except gamma(no-purchase utility) when a particular baseline_product is specified.

Estimating model parameter gamma(no-purchase utility).

Description

Estimating model parameter gamma(no-purchase utility) given estimated model parameters. Internal function that users should not call directly.

Usage

ESTF_gamma(
  eta_wo_k,
  Attr,
  Choice_Set_List,
  Alts_Same_Choice_Set,
  Is_Purchase,
  Base_Prod,
  Num_Prod,
  Num_Attr,
  cond
)

Arguments

Attr

list, a set of data frame in which alternative specific variables.

Choice_Set_List

list, product code included for each choice set.

Alts_Same_Choice_Set

list, for internal calculation, a list of data exposed to the same choice set.

Is_Purchase

list, for internal calculation, list of customers who are exposed to the same choice set who purchased the product. If purchased 1, otherwise 0.

Base_Prod

numeric, the baseline product number

Num_Prod

numeric, the number of all products in observed data.

Num_Attr

numeric, the number of attributes in observed data.

cond

numeric, the value associated with prop, which is the argument of the rmm function. (1-prop)/prop.

Value

The estimated value of parameters except gamma(no-purchase utility).

Data from a Major Hotel Chain

Description

'Hotel_Long', a 'Long format', is a preprocessing data of the publicly available 'Hotel 1' data introduced in Bodea et al. (2009).

Usage

Hotel_Long

Format

'Hotel_Long': A data frame with 8,318 rows and 11 variables:

Booking_ID

ID associated with a booking. Begins at one for each hotel property.

Purchase

Indicator variable equal to one if the product identified by product ID is purchased, zero otherwise.

Room_Type

Code describing the room type associated with the product ID.

Price

The average nightly rate the customer pays in USD (e.g., $199.99). Note that the average nightly rate will not match the rate of any available product rates if an upsell occurs at time of check-in, if the customer requests a specific discount rate at time of check-in, etc.

Party_Size

Number of adults and children associated with the booking.

Membership_Status

Status in rewards program (0—not a member, 1—basic, 2—elevated, 3—premium).

VIP_Membership_Status

Membership status of a VIP rewards program member (0—not a VIP, 1—basic VIP, 2—premium VIP member).

Booking_Date

Date the booking was created (e.g., 20070303 = March 3, 2007).

Check_In_Date

Check-in date (e.g., 20070307 = March 7, 2007).

Check_Out_Date

Check-out date (e.g., 20070310 = March 10, 2007).

Length_of_Stay

Length of stay/number of nights (e.g., three).

Details

'Hotel 1' data contains information on the available alternatives, i.e., choice sets and the associated prices at the time of each customer’s booking decision. We preprocessed 'Hotel 1' data and provide it in two types of data format, 'Hotel_Long' and 'Hotel_Wide'.

The following are the preprocessing of 'Hotel 1' data.

1. Customers' booking transactions that had only one room type available in their choice set were removed as our methods require at least two different products in each choice set.

2. Duplicate records was removed.

3. Choice sets with less than 30 observations, representing rare case were removed.

Source

doi:10.1287/msom.1080.0231

Data from a Major Hotel Chain

Description

'Hotel_Wide', a 'Wide format', is a preprocessing data of the publicly available 'Hotel 1' data introduced in Bodea et al. (2009).

Usage

Hotel_Wide

Format

'Hotel_Wide': A data frame with 1,100 rows and 22 variables:

Booking_ID

ID associated with a booking. Begins at one for each hotel property.

Party_Size

Number of adults and children associated with the booking.

Membership_Status

Status in rewards program (0—not a member, 1—basic, 2—elevated, 3—premium).

VIP_Membership_Status

Membership status of a VIP rewards program member (0—not a VIP, 1—basic VIP, 2—premium VIP member).

Booking_Date

Date the booking was created (e.g., 20070303 = March 3, 2007).

Check_In_Date

Check-in date (e.g., 20070307 = March 7, 2007).

Check_Out_Date

Check-out date (e.g., 20070310 = March 10, 2007).

Length_of_Stay

Length of stay/number of nights (e.g., three).

Room_Type

Code describing the room type associated with the product ID.

Purchased_Prod_Code

Unique number of products associated with room type.

Exposed_Choice_Set

Choice set exposed to individual

Exposed_Choice_Set_Code

Unique number of choice sets associated with Exposed_Choice_Set

Price_1 ~ Price_10

The average nightly rate the customer pays in USD (e.g., $199.99). Note that the average nightly rate will not match the rate of any available product rates if an upsell occurs at time of check-in, if the customer requests a specific discount rate at time of check-in, etc.

Details

'Hotel 1' data contains information on the available alternatives, i.e., choice sets and the associated prices at the time of each customer’s booking decision. We preprocessed 'Hotel 1' data and provide it in two types of data format, 'Hotel_Long' and 'Hotel_Wide'.

The following are the preprocessing of 'Hotel 1' data.

1. Customers' booking transactions that had only one room type available in their choice set were removed as our methods require at least two different products in each choice set.

2. Duplicate records was removed.

3. Choice sets with less than 30 observations, representing rare case were removed.

Source

doi:10.1287/msom.1080.0231

Calculating standard error of estimate.

Description

Calculating statndard error of model parameters and gamma(no-purchase utility). Internal function that users should not call directly.

Usage

Sandwich(
  eta_wo_k,
  gamma,
  Attr,
  Choice_Set_List,
  Alts_Same_Choice_Set,
  Is_Purchase,
  Base_Prod,
  Num_Prod,
  Num_Attr,
  H,
  cond
)

Arguments

eta_wo_k

numeric, estimated model parameters normalized for baseline product.

gamma

numeric, estimated gamma(no-purchase utility).

Attr

list, a set of data frame in which alternative specific variables.

Choice_Set_List

list, product code included for each choice set.

Alts_Same_Choice_Set

list, for internal calculation, a list of data exposed to the same choice set.

Is_Purchase

list, for internal calculation, list of customers who are exposed to the same choice set who purchased the product. If purchased 1, otherwise 0.

Base_Prod

numeric, the baseline product number

Num_Prod

numeric, the number of all products in observed data.

Num_Attr

numeric, the number of attributes in observed data.

H

matrix, a hessian matrix of the log-likelihood function.

cond

numeric, the value associated with prop, which is the argument of the rmm function. (1-prop)/prop.

Value

a value of standard error of estimate.

Calculating a value of score function with respect to parameter gamma.

Description

Calculating a value of score function with respect to parameter gamma. Internal function that users should not call directly.

Usage

U(
  eta_wo_k,
  gamma,
  Attr,
  Choice_Set_List,
  Alts_Same_Choice_Set,
  Is_Purchase,
  Base_Prod,
  Num_Prod,
  Num_Attr,
  cond
)

Arguments

eta_wo_k

numeric, estimated model parameters normalized for baseline product.

gamma

numeric, estimated gamma(no-purchase utility).

Attr

list, a set of data frame in which alternative specific variables.

Choice_Set_List

list, product code included for each choice set.

Alts_Same_Choice_Set

list, for internal calculation, a list of data exposed to the same choice set.

Is_Purchase

list, for internal calculation, list of customers who are exposed to the same choice set who purchased the product. If purchased 1, otherwise 0.

Base_Prod

numeric, the baseline product number

Num_Prod

numeric, the number of all products in observed data.

Num_Attr

numeric, the number of attributes in observed data.

cond

numeric, the value associated with prop, which is the argument of the rmm function. (1-prop)/prop.

Value

a value of score function with respect to parameter gamma.

Calculating a value of score function with respect to parameter gamma for each individual and each alternative.

Description

Calculating a value of score function with respect to parameter gamma for each individual and each alternative. Internal function that users should not call directly.

Usage

iU(
  eta_wo_k,
  gamma,
  Attr,
  Choice_Set_List,
  Alts_Same_Choice_Set,
  Is_Purchase,
  Base_Prod,
  Num_Prod,
  Num_Attr,
  cond
)

Arguments

eta_wo_k

numeric, estimated model parameters normalized for baseline product.

gamma

numeric, estimated gamma(no-purchase utility).

Attr

list, a set of data frame in which alternative specific variables.

Choice_Set_List

list, product code included for each choice set.

Alts_Same_Choice_Set

list, for internal calculation, a list of data exposed to the same choice set.

Is_Purchase

list, for internal calculation, list of customers who are exposed to the same choice set who purchased the product. If purchased 1, otherwise 0.

Base_Prod

numeric, the baseline product number

Num_Prod

numeric, the number of all products in observed data.

Num_Attr

numeric, the number of attributes in observed data.

cond

numeric, the value associated with prop, which is the argument of the rmm function. (1-prop)/prop.

Value

a value of score function with respect to parameter gamma for each individual and each alternative.

The log-likelihood value for each individual and each alternative.

Description

For function that calculating standard error of estimate. Internal function that users should not call directly.

Usage

iloglike(
  eta,
  Choice_Set_List,
  Choice_Set_Code,
  Base_Prod,
  Num_Prod,
  Num_Attr,
  Alts_Same_Choice_Set,
  Is_Purchase
)

Arguments

eta

numeric vector, the model parameters to be estimated

Choice_Set_List

list, product code included for each choice set.

Choice_Set_Code

numeric, choice set code Number specified for calculating log-likelihood.

Base_Prod

numeric, the baseline product number

Num_Prod

numeric, the number of all products in observed data.

Num_Attr

numeric, the number of attributes in observed data.

Alts_Same_Choice_Set

list, for internal calculation, a list of data exposed to the same choice set.

Is_Purchase

list, for internal calculation, list of customers who are exposed to the same choice set who purchased the product. If purchased 1, otherwise 0.

Value

The log-likelihood value for each individual and each alternative.

Calculating the log-likelihood function.

Description

Calculating model's log-likelihood function. Internal function that users should not call directly.

Usage

loglike(
  eta,
  Choice_Set_List,
  Choice_Set_Code,
  Base_Prod,
  Num_Prod,
  Num_Attr,
  Alts_Same_Choice_Set,
  Is_Purchase
)

Arguments

eta

numeric vector, the model parameters to be estimated

Choice_Set_List

list, product code included for each choice set.

Choice_Set_Code

numeric, choice set code Number specified for calculating log-likelihood.

Base_Prod

numeric, the baseline product number

Num_Prod

numeric, the number of all products in observed data.

Num_Attr

numeric, the number of attributes in observed data.

Alts_Same_Choice_Set

list, for internal calculation, a list of data exposed to the same choice set.

Is_Purchase

list, for internal calculation, list of customers who are exposed to the same choice set who purchased the product. If purchased 1, otherwise 0.

Value

The value of log-likelihood.

Predict method for Revenue Management Model Fits

Description

Predicted values based on RMM object

Usage

## S3 method for class 'rmm'
predict(object, newdata, Rem_Choice_Set, Choice_Set_Code, fixed = TRUE, ...)

Arguments

object

Object of class inheriting from "rmm"

newdata

A data frame in which to look for variables with which to predict.

Rem_Choice_Set

List of choice sets remaining in the data.

Choice_Set_Code

Specifies the choice set of newdata.

fixed

If fixed=TRUE, the alternative with the highest prediction probability is determined as decision. Otherwise (fixed=FALSE), one of the alternatives is determined in proportion to the predictive probability.

...

further arguments passed to or from other methods.

Value

preict.rmm produces a list of predictions, which contains decisions and probabilities.

Examples


data(Hotel_Long)

# Before using the rmm function, the user must first use the rmm_shape function.
rst_reshape <-  rmm_reshape(data=Hotel_Long, idvar="Booking_ID",
     alts="Room_Type", asv="Price", resp="Purchase", min_obs=30)

# Fitting a model
rst_rmm <- rmm(rst_reshape, prop=0.7, model="cl")

# Predictions
Rem_Choice_Set <- rst_reshape$Rem_Choice_Set

newdata1 <- data.frame(Price_1=c(232, 122, 524), Price_3=c(152, 531, 221),
                       Price_4=c(163, 743, 192), Price_5=c(132, 535, 325),
                       Price_7=c(136, 276, 673), Price_8=c(387, 153, 454),
                       Price_9=c(262, 163, 326), Price_10=c(421, 573, 472))

predict(rst_rmm, newdata=newdata1, Rem_Choice_Set=Rem_Choice_Set,
        Choice_Set_Code=3, fixed=TRUE)


newdata2 <- data.frame(Price_1=c(521, 321, 101, 234, 743),
                       Price_5=c(677, 412, 98, 321, 382),
                       Price_8=c(232, 384, 330, 590, 280))

predict(rst_rmm, newdata=newdata2, Rem_Choice_Set=Rem_Choice_Set,
        Choice_Set_Code=7, fixed=FALSE)

Fitting Revenue Management Models

Description

rmm is used to fit Revenue Management Models. Users can specify cl (conditional logit model) and ml (multinomial logit model) as RMM model.

Usage

rmm(rmm_data, prop = 0.7, model = "cl")

Arguments

rmm_data

an object of class "rmm_data", a output of rmm_reshape function.

prop

numeric, user assumed market share.

model

character, specify fitting method ("cl" or "ml"). "cl" (default) refers to the Conditional Logit Model, and "ml" refers to the Multinomial Logit Model.

Value

rmm returns an object of class inheriting from "rmm".

See Also

rmm fits the model with the RDE method introduced in doi:10.2139/ssrn.3598259.

Examples


data(Hotel_Long)

# Before using the rmm function, the user must first use the rmm_shape function.
rst_reshape <- rmm_reshape(data=Hotel_Long, idvar="Booking_ID", alts="Room_Type",
                           asv="Price", resp="Purchase", min_obs=30)

# Fitting a model
rst_rmm <- rmm(rst_reshape, prop=0.7, model="cl")
print(rst_rmm)

Reshape Long-Format Data

Description

This function reshapes a 'Long-Format' data (with the repeated measurements in separate rows) to 'Wide-Format' data (with repeated measurements in separate columns of the same row). The reshaped 'wide-format' data is an S3 class called 'rmm_data' and contains information for fitting the model with the rmm function. Users who want to use the rmm function must first use the rmm_reshape function. The rmm function receives only S3 class 'rmm_data' as input.

Usage

rmm_reshape(data, idvar, resp, alts, asv, min_obs)

Arguments

data

data frame, a 'Long-Format' transaction data.

idvar

character, variable name representing each individual's id in the transaction data.

resp

character, variable name representing result of a individual choice.

alts

character vector, variable names representing a alternatives.

asv

character vector, variable names representing a alternative specific variables.

min_obs

numeric, specify the minimum observation for each choice set in the transaction data.

Value

The 'Wide-Format' data and various information required for the rmm function.

See Also

rmm for estimating parameters.

Examples


data(Hotel_Long)

rst_reshape <- rmm_reshape(data=Hotel_Long, idvar="Booking_ID",
resp="Purchase", alts="Room_Type", asv="Price", min_obs=30)

class(rst_reshape)  # "rmm_data"
ls(rst_reshape)     # "Alts_Code_Desc" "ASV" "asv_name" "data_wide"
                    # "Rem_Choice_Set"     "Removed_Choice_Set"

rst_reshape$data_wide  # reshaped data