The Falling Weight Deflectometer (FWD), among many other non-destructive analysis techniques, is vital to evaluating the structural condition of roadways and optimizing pavement management systems. Various methods based on pavement surface deflection, as measured by FWD, are widely used for assessing the structural stability of pavements across the globe. However, performing these tests consistently at the network-level is challenging, and data interpretation requires time, technical expertise, finances, and other resources. As a result, the structural component of roadways is often neglected when making decisions about maintenance and repair. This study proposes a machine learning approach, specifically the random forest algorithm, to estimate four basic deflection basin parameters, including surface curvature index, Base Damage Index, Base Curvature Index, and Deflection Ratio, using various input parameters such as structural, functional, environmental, and subgrade soil properties. To develop an effective model, the random forest algorithm is trained and tested using MATLAB tool, using the data gathered through field trials. The prediction accuracy from the developed model is evaluated based on the root mean square error and coefficient of determination value. Based on the input variables, the random forest method constructs numerous decision trees, each of which offers a prediction for the deflection basin parameters. The algorithm then combines the predictions of all the trees to obtain a final prediction. This approach improves the accuracy of deflection basin parameter prediction compared to traditional methods, which rely on backcalculation of pavement layer moduli. Overall, the proposed machine learning approach using the random forest algorithm provides an efficient and accurate way to predict the deflection basin parameters of flexible pavements.