Enhanced Sales Forecasting Using Machine Learning: Integrating Customer Sentiment and Market Trend Analysis

Sales forecasting is a vital business tool aimed at fine-tuning pricing, improving market positioning, and maximizing inventory levels. This research proposes a hybrid model that combines consumer behavior insights with data-driven methods to enhance the accuracy of sales predictions. Sales patterns are evaluated for this using complex machine learning and statistical techniques such Random Forest (RF), Linear Regression (LR), Integrated Autoregressive Moving Average (ARIMA), Long Short-Term Memory (LSTM), and Extreme Gradient Boosting (XGI). To determine the dependability of the prediction, the models are evaluated against important performance criteria such Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). Empirical evidence shows that Random Forest model is most accurate with the lowest RMSE of $\mathbf{1 9. 2 8 \%}$ and MAE of 8.23%. Linear Regression and LSTM both display high prediction errors with an RMSE value of $\mathbf{1 3 4. 9 2 \%}$ and $\mathbf{1 0 8. 4 0 \%}$, respectively, indicating their lower efficacy. ARIMA displays moderate efficacy with an RMSE of $\mathbf{3 9. 8 2 \%}$ and MAE of $\mathbf{3 5. 3 9 \%}$, showing satisfactory predictability. Conversely, XGBoost shows a balanced performance with an RMSE of $\mathbf{4 8. 0 0 \%}$ and MAE of $\mathbf{2 6. 8 \%}$, an option that can be considered for forecasting purposes. The findings demonstrate the need to choose the right forecasting model since differences in performance can reflect heavily on strategic business decisions. Through the use of efficient models such as Random Forest, organizations can improve decisionmaking in market position, inventory, and competitive strategy, leading to better operational efficiency and profitability.