Smart contracts, deployed on blockchain platforms, have revolutionized various industries by enabling trustless and automated transactions. However, these contracts immutable and decentralized nature has made them an attractive target for malicious actors seeking to exploit vulnerabilities. This research proposes a novel approach for vulnerability detection, which extends its focus to a broader feature space encompassing both the source and bytecode elements to enhance the security and robustness of smart contracts. In this paper, we make three significant contributions. First, we amass a dataset comprising nearly twenty-three thousand real-world solidity smart contracts. Second, we create a comprehensive feature space by combining high-level and low-level features. High-level features are extracted from the source code using graph learning techniques for rich control and data semantics analysis, and low-level features are extracted from the opcode sequence. Finally, we employ data-driven methods on a dataset balanced using SMOTETomek. Our results demonstrate improved vulnerability detection compared to state-of-the-art models, with the highest score of 93.9%.