OTAPI: Optimizing Treatment Allocation in the Presence of Interference

This repository provides the code for the paper "OTAPI: Optimizing Treatment Allocation in the Presence of Interference".

The structure of the code is as follows:

OTAPI/
|_ data/
  |_ semi_synthetic/                   
    |_ BC/
    |_ Flickr/
|_ scripts/
  |_ BC_experiment.py                    
  |_ Enron_experiment.py                    
  |_ Flickr_experiment.py
  |_ runtime_experiment.py           #Measure runtimes of the different methods  
  |_ spillover0_experiment.py        #Experiment with spillover effect == 0    
  |_ spillover0.1_experiment.py      # "" == 0.1
  |_ spillover0.3_experiment.py      # "" == 0.3
  |_ spillover0.5_experiment.py      # "" == 0.5
  |_ spillover0.7_experiment.py      # "" == 0.7
  |_ spillover_figures.py            # Generate figures after all spillover experiments 
  |_ watts_strogatz.py
|_ src/
  |_ data/
    |_ data_generator.py                 # Code to generate synthetic and semi-synthetic data
    |_ datatools.py                 
  |_ methods/
    |_ allocation/         # Allocation optimization
      |_ utils/
        |_ CELF.py                    # Code for CELF algorithm
        |_ Genetic_algorithm.py       # Code for genetic algorithms
        |_ allocation_utils.py        # Utils for the allocations
      |_ allocations_runtime.py     # Code to measure the runtime of the allocation methods
      |_ extra_allocations.py       # Code to run extra experiments for the appendix
      |_ get_allocations.py         # Code to run and save the allocations
    |_ causal_models/         # Train relational causal estimator
      |_ baselineModels.py          # Baseline models (e.g., TARNet and CFR)
      |_ model_training.py          # Model training and evaluation
      |_ layers.py                  
      |_ model.py                   # NetEst model
      |_ modules.py                 
      |_ model_tuning.py            # Train best model
    |_ utils/
      |_ plotting.py/      # Generate figures
      |_ utils.py/

Installation.

The requirements.txt provides the necessary packages. All code was written for python 3.10.13.

Usage

Download the data for the BC, Enron, and Flickr datasets from Google Drive. The original Flickr and BC data comes from this repo. We use the same data as Jiang & Sun (2022). The orginal Enron dataset comes from Network Repository.

Put the data in the data/semi_synthetic/ folder. Now, the results from the paper can be reproduced by setting the DIR variable to your directory and running the appropriate script. To reproduce Figures 5a and 5b, the results from all the spillover*_experiment.py are needed before running the spillover_figures.py file to generate the figures.

Acknowledgements

Our code builds upon the code from Jiang & Sun (2022).

Jiang, S. & Sun, Y. (2022). Estimating causal effects on networked observational data via representation learning. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management, (pp. 852–861).