Manufacturing industries face significant challenges in maintaining consistent product quality, particularly in minimizing reject rates across production machines, as high reject levels not only increase operational costs but also reduce overall efficiency and competitiveness. This study aims to develop a predictive approach using the Random Forest algorithm to forecast monthly chip rejects across different production machines, with historical reject data consisting of 1,820 records from June 2023 to September 2024 analyzed based on four primary reject categories and five production machines (DCL1, DCL2, CMI200, CMI200+, and YMJ400). The Random Forest model was applied to classify and predict reject patterns, and its performance was evaluated based on prediction accuracy and error rates, showing that the algorithm is effective in predicting reject counts with an absolute error of 0.640 ± 0.183, especially for lower reject values under 300, although accuracy decreases when handling higher reject levels above 500. Machine-level analysis further reveals that DCL1 and DCL2 consistently contribute the highest reject counts with high variability, while CMI200 and CMI200+ demonstrate stable performance with most rejects below 300, and YMJ400 generally records lower rejects but occasionally exhibits spikes, suggesting inconsistent performance. In conclusion, the Random Forest model provides a reliable predictive framework for monitoring reject trends, identifying DCL1 and DCL2 as priority targets for improvement, and supporting proactive maintenance strategies to enhance overall production quality.