The modified sine distribution and machine learning models for enhancing crude oil production prediction

Economic data modeling and forecasting are crucial for facilitating decision-making in the industrial and financial realms. This paper describes a dual-objective study focusing on modeling and forecasting economic data. The Modified Sine (MoS) distribution is introduced in the modeling phase, which was constructed utilizing an innovative transformation technique. This technique incorporates an additional parameter, enhancing the distribution's flexibility. The MoS distribution exhibits diverse density shapes; its hazard function is versatile, displaying decreasing and inverted bathtub patterns. These features make the MoS distribution highly suitable for analyzing economic and industrial datasets. Key statistical properties have been identified, and parameters are evaluated using maximum likelihood estimation (MLE). The MoS distribution's performance is validated by applying it to crude oil production data, where it outperforms established models in goodness-of-fit metrics. Machine learning models are employed to forecast economic trends in the prediction phase. Ridge, Support Vector Regression (SVR), Linear Regression, Lasso, and Random Forest are some of the models employed. They are assessed using measures like mean absolute error (MAE), mean absolute percentage error (MAPE), mean squared error (MSE), and root mean squared error (RMSE). Recursive Feature Elimination (RFE) is used to identify significant predictors, reducing model complexity and overfitting. Random Forest achieves the lowest error rates before and after feature selection, consistently demonstrating the best predictive performance. The study reveals that feature selection significantly enhances model accuracy, with Random Forest delivering the most precise predictions. This integrated approach highlights the MoS distribution as a reliable instrument for statistical computing.