Enhancing Obsolescence Forecasting with Deep Generative Data Augmentation: A Semi-Supervised Framework for Low-Data Industrial Applications
Elie Saad, Mariem Besbes, Marc Zolghadri, Victor Czmil, Claude Baron, Vincent Bourgeois
TL;DR
The paper tackles obsolescence forecasting in industries with long asset lifetimes where labeled data are scarce. It introduces a hybrid framework that first uses deep generative models to augment tabular data and then applies a semi-supervised learning algorithm to leverage both labeled and unlabeled data for robust forecasting, decoupling the generative and discriminative components to remain model-agnostic. The approach combines an invertible dimensionality-reduction autoencoder, three generative models (Real NVP, TVAE, CTGAN), and a Random Forest discriminator, and it demonstrates substantial accuracy gains on GSM Arena and Arrow datasets compared with state-of-the-art baselines. Practically, the framework offers a flexible, scalable solution for proactive obsolescence management in industry, enabling high-accuracy forecasting even under extreme data paucity and long life cycles.
Abstract
The challenge of electronic component obsolescence is particularly critical in systems with long life cycles. Various obsolescence management methods are employed to mitigate its impact, with obsolescence forecasting being a highly sought-after and prominent approach. As a result, numerous machine learning-based forecasting methods have been proposed. However, machine learning models require a substantial amount of relevant data to achieve high precision, which is lacking in the current obsolescence landscape in some situations. This work introduces a novel framework for obsolescence forecasting based on deep learning. The proposed framework solves the lack of available data through deep generative modeling, where new obsolescence cases are generated and used to augment the training dataset. The augmented dataset is then used to train a classical machine learning-based obsolescence forecasting model. To train classical forecasting models using augmented datasets, existing classical supervised-learning classifiers are adapted for semi-supervised learning within this framework. The proposed framework demonstrates state-of-the-art results on benchmarking datasets.