Random Forest algorithm, a powerful machine learning technique recognized for its flexibility and high predictive accuracy has been used in different applications across various domains. This paper presents a comparative analysis of the performance of Random Forest in forecasting multivariate time series data exhibiting trend and seasonality, a task that poses a significant challenge since the traditional methods often used to forecast time series can fail to accurately model these complex characteristics, leading to poor forecasting performance and unreliable prediction. In this study, different optimization techniques are used to enhance the performance of the RF models, such as feature engineering, using various features (temporal, lagged, rolling window, and interaction features), and parameter tuning (Manual, Random Search, and Grid Search). The models are evaluated using three accuracy metrics: Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) with lower values indicating better performances, the results are then compared to traditional forecasting methods. The proposed forecasting framework demonstrated the capability of the RF model to capture trends and seasonality in multivariate time series and provide competitive results.