Finding the Minimum Information Partition (MIP) is a computational challenge for the advancement of Integration Information Theory (IIT) due to the enormous number of calculations required, even for systems with relatively few elements. This work introduces HDMP (Heuristic-Driven Memoization Process), a novel algorithmic approach to efficiently calculate high-quality approximations of the MIP for medium-scale systems. Unlike approximation methods, HDMP leverages the mathematical properties of conditional independence and heuristic guidance through cost matrix analysis to efficiently navigate the partition search space. In this study, various approaches were employed in order to improve computational performance. A thorough review of different test cases was conducted, contrasting the results with those obtained using the PyPhi application. HDMP demonstrated over 90% reduction in execution time for systems with 200 nodes, achieving practical feasibility for bipartite systems up to 100 nodes compared to PyPhi’s limit of approximately 30 nodes. The algorithm combines dynamic programming with memoization and uses a heat map-based heuristic to prioritize high-potential partitions. These improvements extend the computational boundaries of IIT, enabling analysis of substantially larger networks while preserving the theoretical guarantees of finding the true minimum information partition.