Predicting contract length with probabilistic programming

Click for: original source

An article by Antoine Hachez about Jobandtalent experience with applying Artificial Intelligence (AI) to real business problems.

One of the key metrics in this industry is obviously the length of the contracts as it is proportional to the revenue per contract, which is compared to the fixed and variable costs of managing it.

The article describe how their data team tackled a challenge:

  • The modeling process
  • Modeling over time
  • Probabilistic programming – the simple model
  • Probabilistic programming – the time-dependent approach
  • Project evolution and conclusion

They describe how their approach is flexible and how they quickly iterated to create the KPIs that would fit our needs (actionable and unbiased by outliers). The Bayesian approach offers many advantages in this regard. Great read!

[Read More]

Tags big-data data-science miscellaneous how-to