Developing an dynamic estimation model based on artificial intelligence with online data

Prediction in industrial engineering has become a critical area of research in today's data-driven world, especially with the increasing importance of sales predictions. Large data sets formed by the interactions of billions of people in the virtual environment provide valuable information about the past and have great potential for making predictions about the future. However, since a significant portion of this data is in an unstructured form, direct processing with traditional prediction methods creates various difficulties, which increases the need for more advanced and flexible analysis techniques. In this context, the thesis study prepared aims to transform the unstructured big data formed in the online environment into structured data by processing and interpreting it through large language models (LLMs) and to integrate this data as an additional feature to classical prediction models. LLMs are deep learning-based models trained on large data sets and play an important role in increasing prediction accuracy by making meaningful inferences from the data accumulated in the virtual environment thanks to their natural language processing capabilities. In this process, prediction performance has been further improved by using various machine learning algorithms such as linear regression, support vector regression, and gradient support vectors. Thus, thanks to artificial intelligence-supported data processing techniques, it has become possible to produce more reliable, consistent, and accurate predictions by going beyond traditional methods; this has enabled more effective use of decision support systems, especially in sales prediction processes, and has significantly strengthened the integration of big data and machine learning methods into industrial processes.