Making Money from What You Know – How to Sell Information?
Authors: Shani Alkoby, Zihe Wang, David Sarne, Pingzhong Tang2421-2428
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | For the case where buyers are human subjects, using an extensive set of experiments we show that the methods result in substantially different outcomes. Furthermore, using standard machine learning techniques the information provider can rather accurately predict the performance of the different methods for new problem settings, hence substantially increase profit. |
| Researcher Affiliation | Academia | 1The University of Texas at Austin, Austin, Texas, USA 2Shanghai University of Finance and Economics, Shanghai, China 3Bar-Ilan University, Ramat Gan, Israel 4Tsinghua University, Beijing, China |
| Pseudocode | No | The paper describes mathematical models and experimental procedures in prose and equations, but it does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper states 'Our implementation of decision trees was done using python 3.6.4.' but does not provide any link or explicit statement about releasing the source code for its methodology. |
| Open Datasets | No | The paper describes generating its own dataset: 'For this purpose we pre-generated a set of 250 core problem settings.' and 'Overall we had 450 subjects playing the game such that each information providing method was used with 150 of them.'. It does not provide access information (link, DOI, citation) for this custom dataset. |
| Dataset Splits | Yes | For building the tree we used k-fold cross validation. In k-fold cross validation the data is being divided into k equal subsets of the data called bins. Each of those bins, whose size is equal to the total data size divided by k, is being used as a testing set in a separate learning experiment where the tree is being built using the remaining k 1 bins (put together as a training set). Searching for the optimal value of k, i.e., the one which minimizes the error, we found that for the Full Information method, the smallest error is being received for k = 5 whereas for the Random Elimination and Options Menu methods it is being received for k = 10. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running the experiments, such as GPU/CPU models or memory specifications. |
| Software Dependencies | No | The paper mentions 'Our implementation of decision trees was done using python 3.6.4.' However, it only specifies the programming language version and does not list any other key software components (e.g., libraries, frameworks, solvers) with their specific version numbers required for reproducibility. |
| Experiment Setup | Yes | We use the Value of Information concept, VoI... We then set fees to be 0.2 VoI, 0.5 VoI, 0.8 VoI, 1.2 VoI, 1.5 VoI, and 1.8 VoI... For deciding how to split the data at each node we used the Gini impurity... For building the tree we used k-fold cross validation... for the Full Information method, the smallest error is being received for k = 5 whereas for the Random Elimination and Options Menu methods it is being received for k = 10. In order to overcome the risk for overfitting we used post-pruning. |