Creativity is often characterized by the capacity to generate novel ideas, explore unconventional approaches, and solve problems through intuition, curiosity, and innovative thinking. Assessing this multifaceted skill is both essential and challenging, especially in educational and game-based environments where creativity drives engagement and learning. In this study, we expanded upon existing research on creativity measurements in digital games to propose a comprehensive model for identifying creative solutions in serious digital puzzle games. This model was adapted to a geometry-focused puzzle game, with solutions manually evaluated across dimensions such as fluency, flexibility, and surprise. Using the resulting labelled dataset, we trained machine learning models to classify creative solutions, achieving promising results. These findings highlight the potential for integrated methods to advance creativity assessment and enhance educational game design, contributing to the development of tools that inspire and nurture creative thinking in learners.