Decision-making in manufacturing is increasingly complex due to the industry's fast-paced and globalized nature, requiring companies to be agile and data-driven. A systematic review of 420 papers (from Scopus and ResearchGate), narrowed to 40 relevant studies, highlights the evolution from linear models like ARIMA to more advanced non-linear methods such as ANNs, RNNs, and particularly LSTM networks, which address the limitations of long-term sequence modeling. Modern manufacturers are leveraging these machine learning techniques to optimize operations, reduce costs, and enhance product quality across key areas like predictive maintenance, supply chain optimization, and demand forecasting. The review underscores how automated supervision systems, integrated with IoT and big data analytics, are enabling real-time equipment monitoring and early fault prediction, thereby improving system reliability and minimizing downtime. As part of Industry 4.0, this shift towards AI-driven automation is reshaping the manufacturing landscape. The findings serve as a valuable guide for companies seeking to adopt machine learning to improve decision-making and maintain competitiveness.