Customer Sentiment Analysis for Demand Forecasting of Electronic Devices Using Machine Learning Techniques

The growing influence of e-commerce and social me- dia has made traditional demand forecasting methods insufficient due to their limited responsiveness to real-time consumer senti- ment. This paper presents a sentiment-driven demand forecasting framework that combines web scraping, natural language pro- cessing, and machine learning. Product reviews from Amazon and tweets are extracted using BeautifulSoup, analyzed using BERT, NLTK, SpaCy, and a custom LSTM model. Sentiment scores are integrated into a pooled machine learning model alongside historical sales and seasonal data. Bayesian inference is then applied to perform sentiment-weighted product allocation across regions. Experimental results demonstrate improved forecasting accuracy, supporting more adaptive inventory and marketing strategies.