Accurate traffic forecasting is crucial for efficient travel planning, congestion management, and urban mobility optimization. Traditional forecasting methods rely heavily on extensive historical datasets, requiring continuous data collection and processing, which can be costly and impractical in certain scenarios. This study introduces a generic congestion forecasting framework that operates without the need for historical traffic data. Instead of relying on continuous monitoring, the proposed approach leverages a machine learning-based model that extracts relevant information from road characteristics to predict congestion levels for specific time slots. Extensive experiments were conducted to evaluate the effectiveness of our approach. With an accuracy of 90%, our framework provides a cost-effective and scalable alternative to conventional solutions by reducing dependence on large-scale data collection. The proposed method can be seamlessly integrated into smart city applications, enabling real-time traffic management and improved decision-making for urban planning.