Article: Multi-objective optimisation of cigarette tobacco leaf blend formulation based on sensory-chemical correlation and machine learning Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.4 pp.32 - 47 Abstract: Cigarette leaf blend formulation design is a core component in determining product sensory quality. This study proposes a multi-objective optimisation method based on sensory-chemical correlations and machine learning. First, key chemical components of leaf blend samples are systematically collected to construct an initial dataset. Subsequently, multivariate statistical methods such as partial least squares regression are employed to identify the key chemical indicators driving sensory quality. Based on this, a machine learning model based on deep learning is established to accurately predict the key chemical indicators and sensory quality scores of the formulation. Finally, sensory quality, key chemical indicators, and raw material costs are set as optimisation objectives to construct a multi-objective optimisation model. The experimental results show that the multi-objective optimisation model constructed by this method generates 152 Pareto optimal solutions, improving sensory quality by 12%, reducing raw material costs by 19%, and increasing chemical stability by 55%. Inderscience Publishers - linking academia, business and industry through research

Title: Multi-objective optimisation of cigarette tobacco leaf blend formulation based on sensory-chemical correlation and machine learning

Authors: Xingliang Li; Weixian Ren; Guangwei Liu

Addresses: Engineer, Research and Development Center, Gansu Tobacco Industrial Co., Ltd., Lanzhou, 730050, China ' Engineer, Research and Development Center, Gansu Tobacco Industrial Co., Ltd., Lanzhou, 730050, China ' Engineer, Research and Development Center, Gansu Tobacco Industrial Co., Ltd., Lanzhou, 730050, China

Abstract: Cigarette leaf blend formulation design is a core component in determining product sensory quality. This study proposes a multi-objective optimisation method based on sensory-chemical correlations and machine learning. First, key chemical components of leaf blend samples are systematically collected to construct an initial dataset. Subsequently, multivariate statistical methods such as partial least squares regression are employed to identify the key chemical indicators driving sensory quality. Based on this, a machine learning model based on deep learning is established to accurately predict the key chemical indicators and sensory quality scores of the formulation. Finally, sensory quality, key chemical indicators, and raw material costs are set as optimisation objectives to construct a multi-objective optimisation model. The experimental results show that the multi-objective optimisation model constructed by this method generates 152 Pareto optimal solutions, improving sensory quality by 12%, reducing raw material costs by 19%, and increasing chemical stability by 55%.

Keywords: cigarette leaf blend formulation; sensory-chemical correlation; machine learning; multi-objective optimisation.

DOI: 10.1504/IJICT.2026.151526

International Journal of Information and Communication Technology, 2026 Vol.27 No.4, pp.32 - 47

Received: 19 Aug 2025
Accepted: 06 Nov 2025
Published online: 04 Feb 2026 *

Title: Multi-objective optimisation of cigarette tobacco leaf blend formulation based on sensory-chemical correlation and machine learning

Keep up-to-date