** SALES_DATA Folder: **

This folder contains the offer sets for each simulation. Each row is a different offer set each column corresponds to a product.  A 1 indicated that the product was offered and 0 indicated that the product was not offered.  There are 10 products in addition to the no purchase option which is the first column. The no purchase option is always offered.

We have 3 different types of offer sets:

SalesData:  Arrivals in every period
SalesDataNoArrival: Censored demand - potential for a no arrival in each period
SalesDataNested: Offer sets are all of the form (i, i+1, i+2, ..,j) -- strictly speaking these subsets are not nested, they are essentially intervals

There are three numbers in names of the data sets.
first number: Ground choice model # 1-10, we used 10 different ground choice models in our computational experiments
second number: Data set # 1-10, there are 5000 rows, the first 2500 rows are used for training (when we need to have 1000 or 1750 customers in the training data set, we use the first 1000 or first 1750 rows), the last 2500 rows are used for testing
third number: number of customer types 50 and 100 although we only put the results in the paper for the 100 customer type simulations


** PURCHASES Folder: **

This folder contains the purchased product for each simulation. Each row is a different offer set each column corresponds to a product.  A 1 indicated that the product was purchased and 0 indicated that the product was not purchased.  There are 10 products in addition to the no purchase option which is the first column.

We have 3 different types of offer sets:

Purchases:  Arrivals in every period
PurchasesDataNoArrival: Censored demand - potential for a no arrival in each period
PurchasesNested: Offer sets are all of the form (i, i+1, i+2, ..,j) -- strictly speaking these subsets are not nested, they are essentially intervals

There are three numbers in names of the data sets.
first number: Ground choice model # 1-10, we used 10 different ground choice models in our computational experiments
second number: Data set # 1-10, there are 5000 rows, the first 2500 rows are used for training (when we need to have 1000 or 1750 customers in the training data set, we use the first 1000 or first 1750 rows), the last 2500 rows are used for testing
third number: number of customer types 50 and 100 although we only put the results in the paper for the 100 customer type simulations


** NONPARAMMODEL Folder: **

Contains the nonparametric models that were used to generate the data.  Each row is a different customer type’s preference list and the column number gives the ranking. The no purchase option is product 1.  The model is the same within each trial.  

There are two number in the names of the files.

first number:  Ground choice model # 1-10, we used 10 different ground choice models in our computational experiments
second number: number of customer types 50 and 100 although we only put the results in the paper for the 100 customer type simulations




