What makes a good recession prediction model?



What makes a good recession prediction model?

One word: Storytelling. To elaborate, observe the side-by-side comparison of the Elastic Net, Weighted Average, and Support Vector Machine predictions, for the 6-month time frame: The Elastic Net and Weighted Average models score better than the SVM (on a weighted log-loss basis) because they predict higher probabilities of recession in general!

Does machine learning improve forecast accuracy and predictive ability?

Generating forecasts from an elaborate set of machine learning techniques significantly improves the predictive ability and accuracy at all forecast horizons in the vast majority of the cases relative to both univariate and multiple Logit and Probit models.

Is it worthwhile to implement a machine learning framework for recession forecasting?

It is clear that all machine learning techniques display superior forecasting ability compared to the benchmark model, indicating that it might be worthwhile to implement a machine learning framework for recession forecasting. Table 5.

What are the best machine learning models to try?

More complex models such as LSTMs and Convolutional Neural Nets can be tried. We can experiment with more complex MLPs by adding additional layers and larger number of units in hidden layers. Other vectorization schemes such as Wordbatch can be experimented with ML models. Regression models like FTRL and FM_FTRL can also be tried.