Abstarct: In this study, an integrated frxamework for evaluation, prediction, and optimization in hole cleaning in slim and non-vertical wells is presented. Due to the limitation of annular space and complex hydraulic conditions, these types of wells are prone to cuttings accumulation and settling, which can lead to increased torque and drag, reduced rate of penetration, and increased non-productive time. Initially, using real drilling data and engineering correlations, a physical model was developed to calculate annular hydraulic indices, cuttings transport indices, pressure loss, and equivalent circulating density, and baxsed on this, a physical risk criterion was defined to identify desirable and undesirable hole cleaning conditions. Then, baxsed on this labeling, a machine learning model of the histogram-baxsed gradient boosting classifier type was trained to predict the probability of undesirable hole cleaning conditions in real time. The stability of the model results was evaluated using Monte Carlo uncertainty analysis and sensitivity analysis, which indicated the stability of decisions against conventional fluctuations in the inputs. Subsequently, the model output was converted into operational recommendations, and constrained optimization was performed for high-risk points to determine the optimal values of control parameters while observing operational constraints. The results indicate that fluid flow rate and rate of penetration play a key role in reducing hole cleaning risk, and applying the proposed settings can reduce cleaning risk in a considerable portion of critical points. The proposed frxamework enables the conversion of drilling data into reliable operational decisions and can be used as a decision-support tool in managing hole cleaning risk in slim wells.