<p>The present study explored free-forced convection inside an enclosure with double lid filling with CNT (carbon nanotube)-based nanofluid that includes a discrete heat source placed at the bottom surface. The novelty of this investigation lies in the combined assessment of lid motion direction, localized heat source location using CNT-based nanofluid, along with machine learning-based prediction of thermal characteristics. Parametric simulations have been done by considering three cases that utilized the direction of the moving lid. Furthermore, the non-dimensional heat source’s length (<i>ε</i>) has been varied from 0.25 to 0.75, while its position with optimum length on the bottom surface has been adjusted to the left, middle, and right to evaluate its impact on thermal performance. The volume proportion (<i>φ</i>) of nanofluid has been considered from 0.01 to 0.05. The findings have been reported in terms of evaluating Nusselt number (Nu) and bulk fluid temperature evaluation. The findings show that increasing the <i>φ</i> value results in better heat transfer, but increasing the ε value lowers the Nu. Furthermore, with the heat source positioned at the right, case 3 has demonstrated the optimum performance with achieving Nu of 57.58, outperforming cases 1 and 2 by 5% and 33%, respectively, at Gr = 10⁴, <i>φ</i> = 0.05 (SWCNT–water nanofluid), and <i>ε</i> = 0.25. The Nusselt number and temperature are furthered predicted by machine learning models: XGBR, ANN, and MLR. Among them, XGBR has exhibited superior accuracy for predicting the key findings. Additionally, KFold cross-validation has been conducted to find the mean CV score (<i>k</i> = 5). To confirm the performance evaluation and cross-validation, hyperparameter sensitivity analysis has been executed as well.</p>