Research on Time-Series Data Mining focuses on developing advanced techniques to analyze, model, and forecast temporal data across diverse domains. Recent studies highlight deep learning approaches such as Future-Guided Learning (FGL), which uses dynamic feedback mechanisms inspired by predictive coding to enhance long-term forecasting accuracy. Foundation models with few-shot learning capabilities allow efficient adaptation to new time-series tasks without extensive retraining, while systematic reviews provide insights into benchmark datasets, predictive maintenance applications, and commonly used algorithms. Overall, these innovations enable more accurate, scalable, and efficient analysis of time-dependent data, supporting improved decision-making and predictive analytics in complex real-world scenarios.