Assemblies of plasmonic nanoparticles have been proposed for various applications, including photothermal therapy, exploiting surface plasmon coupling phenomena. However, the rational design of fractal nanoparticle assembly remains challenging due to the lack of structural characterizations and modelization of real systems. Here we used quantitative structure property relationship (QSPR), driven by experimental data and statistical analysis, to establish a relationship between structural descriptors of fractal gold nanoparticle (GNP) aggregates and their light-to-heat conversion. 160 assemblies of various size spherical GNP with different polyelectrolyte chains were synthesized, 1 which differ in their global charge, size, mass fractal dimension and plasmonic properties. 15 independent descriptors of structure and properties were extracted and further analyzed by QSPR. Principal component analysis and multilinear regression reveal that light-to-heat conversion is mainly governed by the structure of the aggregates and not by the characteristics of the building blocks. This highlights the key role of the fractal dimension of the aggregate and of the ratio of GNP/polyelectrolyte mass to optimize photothermal effects. Rational criteria to optimize light-to-heat conversion within non-ideal fractal assemblies of GNP were identified, relaxing on the choice of other parameters, such as GNP or aggregate size, that can be adapted to the desired biomedical applications.