Abstarct: Drilling deep wells under high-pressure and high-temperature (HPHT) conditions is among the most costly operations in the oil and gas industry due to its technical complexity, safety constraints, accelerated equipment degradation, and substantial resource consumption. This study aimed to develop a data-driven frxamework for predicting and optimizing drilling costs under such demanding conditions. The proposed frxamework investigated the effects of several operational variables, including bit type, rate of penetration, reservoir pressure and temperature, drilling-fluid type, flow rate, daily consumable-material cost, and hydraulic parameters. Following data collection and preprocessing, the dataset was divided into training and testing subsets, and Random Forest Regression and Multilxayer Perceptron models were employed to identify the nonlinear relationships between operational parameters and the drilling-cost index. Model performance was evaluated using the coefficient of determination, mean absolute error, and root mean squared error. The results demonstrated that both models were capable of predicting variations in drilling costs with satisfactory accuracy. The Multilxayer Perceptron model achieved a test coefficient of determination of 0.9832; however, its perfect training coefficient of determination indicated a potential risk of overfitting and the need for further stability assessment. In comparison, the Random Forest model achieved coefficients of determination of 0.94885 and 0.86887 for the training and testing datasets, respectively, indicating a more balanced relationship between model fitting and generalization. The optimization results identified a PDC bit, a rate of penetration of approximately 23.43 m/h, a reservoir pressure of 490.21 bar, a reservoir temperature of 233.48°C, a synthetic-baxsed drilling fluid, and a flow rate of 21.71 L/s as one of the favorable operational scenarios. Accordingly, integrating machine-learning models with optimization algorithms can provide an effective decision-support frxamework for reducing drilling costs and operational time, improving the selection of drilling parameters, and enhancing the safety and sustainability of deep and HPHT well operations.