Traditional energy management systems mostly rely on static schedules and rule based approaches, often struggle to adapt dynamic factors such as fluctuation in tariffs, renewable energy variability and real-time demands changes. Energy efficacy is becoming high due to increase in demand, environmental factors, and integration of renewable resources. The conventional system fails to adapt dynamic situations such as real-time pricing and unpredictable weather changes. This paper presents a novel machine learning- based framework for energy consumption optimization, aiming to overcome limitations of traditional rule-based energy management systems. It combines predictive modeling (Gradient Boosting, Random Forest) and reinforcement learning (RL), including deep RL (DRL), to forecast consumption patterns and dynamically adjust appliance schedules. Exploratory data analysis (EDA) uncovers key consumption drivers, including temporal and weather-related variables. The system demonstrates a 25% improvement in energy savings and a 20% cost reduction under real- world scenarios like dynamic pricing and extreme weather. A forecasting tool achieves 87% accuracy, and a real-time dashboard enhances usability. The framework is scalable across residential, industrial, and smart grid applications.