The Automated Venture Capitalist

Investing is hard because there are an incredible number of factors that influence to the success or failure of ventures. Some of these factors are within a venture's control, and others are not. This respository contains an AI method, tool, and dataset to understand these factors and the complex interactions between them so that organizations and members of the general public can do a better job assessing risk and identifying investment opportunities.

In this Repository

Specifically, this repository supports information presented in the AAAI-KDF'20 paper "The Automated Venture Capitalist: Data and Methods to Predict the Fate of Startup Ventures".

• Analysis.m is a Matlab script that contains the analyses presented in the paper.

• data/

  • individualData.csv contains information at the individual-level.
  • teamData.csv contains information at the team-level (which was featurized from the individual-level data).

• results/

  • results.csv contains all model performance with leave-one-team-out cross-validation (Tables S4 & S5 in paper).
  • comparison.csv contains classification performance comparisons between crowd, judges, and models (Table 2 in paper).

• modelCoeff/

  • {nomination,success}Model_coeff.csv contains logistic regression model coefficients (Table 3 in paper).
  • {nomination,success}Model_perf.csvcontains logistic regression model performance when trained on all data.

• figures/

  • calibrationPlot_success.png shows model calibration (Figure 1 in paper).
  • costPlot_{nomination,success}.png shows gain/loss of model predictions (Figure 2 in paper).

• functions/ contains supporting scripts.

Citation

@inproceedings{gham2020autovc,
  title={The Automated Venture Capitalist: 
            Data and Methods to Predict the Fate of Startup Ventures},
  author={Ghassemi, Mohammad M. and Song, Christopher and Alhanai, Tuka},
  booktitle={KDF at the Thirty-Fourth AAAI Conference on Artificial Intelligence},
  year={2020}
}

Contact

If you find this interesting and would like to learn more, please contact [email protected]