The increasing sophistication of ransomware attacks on computer systems presents a significant cybersecurity challenge. This paper evaluates the application of digital forensics methodology in ransomware analysis for Windows-based desktops. We propose a forensically sound dynamic analysis framework integrating key digital forensics principles, including evidence integrity and chain of custody, and processes for ransomware investigation. The methodology involves executing ransomware samples in a sandbox environment, collecting system artefacts, and analysing behavioural patterns. To assess the practicality of the proposed approach, machine learning models are trained on forensic-based features extracted from ransomware activity. Experimental results on seven types of models demonstrate that Decision Tree and XGBoost ones achieve the highest accuracy (93%), validating the effectiveness of our forensic-driven approach. These findings suggest digital forensics can enhance ransomware detection, investigation, and mitigation.