Latent Variable Copula Inference for Bundle Pricing from Retail Transaction Data
Authors: Benjamin Letham, Wei Sun, Anshul Sheopuri
ICML 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Simulations and data experiments demonstrate consistency, scalability, and the importance of incorporating correlations in the joint distribution. |
| Researcher Affiliation | Collaboration | Benjamin Letham BLETHAM@MIT.EDU Operations Research Center, Massachusetts Institute of Technology, 77 Mass. Ave., Cambridge, MA 02139 USA Wei Sun SUNW@US.IBM.COM Anshul Sheopuri SHEOPURI@US.IBM.COM IBM Thomas J. Watson Research Center, 1101 Kitchawan Road, Yorktown Heights, NY 10598 USA |
| Pseudocode | No | The paper does not contain any pseudocode or clearly labeled algorithm blocks. |
| Open Source Code | No | The paper does not provide any information about open-source code availability for the methodology described. |
| Open Datasets | Yes | We use the publicly available Ta-Feng dataset, which contains four months of transaction level data from a Taiwanese warehouse club, totaling about 120,000 transactions and 24,000 items (Hsu et al., 2004). |
| Dataset Splits | Yes | We evaluated the predictive performance of the copula model using 10-fold cross validation, by ļ¬tting the model to 9 folds of the data and then evaluating the (predictive) log-likelihood on the remaining fold. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running the experiments. |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers used for the experiments. |
| Experiment Setup | Yes | The parameter vmin was chosen from a uniform distribution over [ 25, 75] and vmax chosen from a uniform distribution over [100, 200]. For all simulations, the transactions were spread uniformly across three price points, with the prices of the two items taken to be 100 for one third of transactions, 75 for one third, and 50 for the remaining third. For these results we took the item prices as the mode of the price distribution in the data, and since the item costs are unknown, we set them to a 35% markup. |