Enhancing Decision-Making in New Product Development: Forecasting technologies revenues using a Multidimensional Neural Network

Marie Saade (1) , Maroun Jneid (2) , Imad Saleh (1)

1 PARAGRAPHE - Laboratoire Paragraphe
2 UA - Université Antonine

Aiming to retain their position in the marketplace, organizations are constantly enhancing research and development-based digital innovation activities in order to constantly develop new products and deploy new technologies. However, innovative trends and products are prone to failure, leading to undesired repercussions. In addition, when evaluating a product lifecycle, many decision-makers confront unprecedented challenges related to the estimation of potential disruptive innovation. To address this gap and to tackle the opportunities of digitalization, we conduct quantitative study to investigate the usage of research and development activities that can represent a main economic driver for new product/service development. A new approach for predicting innovative technology-based product success is proposed using Neural Networks models and based on the analysis of patents, publications and technologies revenues which are considered major key performance indicators in measuring technology-based product power. The proposed methodology consists of two main steps: forecasting patents and publications growths separately for a specific candidate technology using a common predictive Neural Network regression model, then integrating the results into a Multidimensional Neural Network classifier model in order to predict future revenue growth for this candidate technology. The present methodology is applied using two different types of Neural Networks for comparison purpose: "Wide and Deep Neural Networks" and "Recurrent Neural Networks". Consequently, addressing this estimation represents a decision support and a crucial prerequisite step before proceeding with investments, where organizations can improve decision making in innovative technology-based product/service development. The findings show that the Recurrent Neural Networks models achieve higher prediction accuracy, and outperform the Wide and Deep Neural Networks, proving to be a more reliable model that can enhance digital innovation development.

Mots clés: Innovative product development; Forecasting product revenues; Neural networks; Decision support; Information systems; Analysis of patents; Patents Data

Domaines: Sciences de l'Homme et Société; Informatique [cs]

Marie Saade, Maroun Jneid, Imad Saleh. Enhancing Decision-Making in New Product Development: Forecasting technologies revenues using a Multidimensional Neural Network. EMCIS 202017th European Mediterranean & Middle Eastern Conference on Information Systems, 2020, Dubai, United Arab Emirates. ⟨hal-03017387⟩