Complexity of Products: The Effect of Data Regularisation

Area of Science:

• Economic Complexity
• Statistical Learning
• Econometrics

Background:

• Economic Complexity (EC) field advancements include new quantitative forecasting and data denoising techniques.
• Existing EC literature often uses noisy datasets, impacting analysis.
• New algorithms like Bootstrapped Selective Predictability Scheme (SPSb) and Hidden Markov Model (HMM) regularization have emerged.

Purpose of the Study:

• To analyze the relationship between SPSb and Nadaraya-Watson Kernel regression.
• To investigate the impact of HMM regularization on Product Complexity and logPRODY metrics.
• To explore novel effects of HMM regularization on EC datasets.

Main Methods:

• Proving the convergence of SPSb to Nadaraya-Watson Kernel regression.
• Applying HMM regularization to Product Complexity and logPRODY metrics.
• Analyzing noise reduction and network properties in export data.

Main Results:

• SPSb is interchangeable with Nadaraya-Watson Kernel regression, offering lower time complexity and deterministic outputs.
• HMM regularization confirms the logPRODY model's interpretation of global market structure changes and provides a new interpretation for the Complexity measure.
• HMM regularization reduces data noise and increases nestedness in the export network adjacency matrix.

Conclusions:

• Nadaraya-Watson Kernel regression offers a computationally efficient alternative to SPSb for economic forecasting.
• HMM regularization enhances the understanding of product market dynamics and complexity.
• Regularization techniques are valuable for improving the quality and interpretability of Economic Complexity data.