This research aims to design and optimize a product recommendation system in the Iranian e-commerce context. This research is applied in terms of purpose, the data is numerical and discrete. The aim of this study is to increase understanding of how variables such as customer preferences, product characteristics, and market dynamics affect packaging strategies in the Iranian market. The results indicate that the proposed model is a new path for personalizing product group offerings and marketing managers can use it in their decision-making. Data analysis was performed using coding in MATLAB software and the model outputs are presented with the proof of the relevant formulas. Evaluations show that companies can determine product recommendations more accurately by applying the proposed model. For this purpose, self-organizing neural network (SOM) and support vector machine (SVM) algorithm were used to cluster data and predict user preferences. The research data was extracted from the Digikala store and after normalization with Z-score, it was examined in the form of different scenarios based on influential variables including age, gender, education, product type and customer satisfaction. The results showed that the presented hybrid model is able to correctly predict product suggestions with an accuracy of 94%, which is better than the comparative methods SOM and SVM with an accuracy of 59% and 48%, respectively. Using the silhouette index to measure the quality of clustering showed that the optimal number of clusters was 7 and the largest group related to the food products category with a share of 45%.