Presenting a Model Based on the Use of Artificial Intelligence Algorithms in Marketing for Small and Medium-Sized Enterprises in the Food Industry

Abstract

In the competitive world of the food industry, small and medium-sized enterprises (SMEs) face marketing challenges such as resource limitations, intense competition, and the need for personalization. Artificial intelligence (AI) technology, with its capability to analyze big data and predict customer behavior, holds transformative potential. However, the lack of localized frameworks for its implementation in these businesses has created a significant gap. The objective of this research is to present a comprehensive model based on XGBoost and DistilBERT algorithms to optimize marketing strategies for SMEs in the food industry. Key research questions include identifying factors influencing AI adoption, modeling them using XGBoost, analyzing customer behavior with DistilBERT, and comparing model accuracy.

The research method is mixed (qualitative-quantitative) with an interpretivist and inductive approach. In the qualitative section, thematic analysis using the Attride-Stirling method was applied to semi-structured interviews with 12 experts (managers and specialists) to extract 25 organizing themes (such as technological infrastructure, human resources, and managerial factors). In the quantitative section, a Likert scale questionnaire was distributed to 384 managers/specialists. Data normality (Sig > 0.05), reliability (α > 0.78), and descriptive analysis (means 3.45–4.25) were confirmed. Modeling was conducted using XGBoost for numerical factors and DistilBERT for textual data.

Findings revealed key AI adoption factors: managerial acceptance (mean 4.25), personalization (4.25), and financial resources (3.65). The XGBoost model, with 92% accuracy, R² = 0.85, and RMSE = 1.23, prioritized factors (managerial acceptance: gain = 15.2). The DistilBERT model, with 91% accuracy and F1 = 0.87, analyzed customer behavior from textual data (acceptance: attention = 0.65). DistilBERT outperformed in low-error cases (145 samples < 0.3).

The proposed model, integrating XGBoost (for numerical prediction) and DistilBERT (for textual analysis), provides a localized framework enabling SMEs to design targeted marketing strategies, reduce costs by 15–25%, and increase conversion rates. Recommendations include managerial training, data investment, and collaboration with AI startups. This model addresses the local gap and supports the competitive growth of SMEs in the dynamic food industry market.