This document presents the choice and the optimization of a supervised classification algorithm, applied on plantar pressure measurements. Its purpose is to classify, based on pressure data coming from a smart insole, whether the user has risks of foot ulceration or not. The end goal is to implement this algorithm in a smart insole developed at LAI. The insole is able to perform weight redistribution to correct the abnormal plantar pressures measured on diabetic patients using actuators based on magneto-rheological fluids. Different algorithms are trained, optimized and evaluated using a dataset containing plantar pressure measurements of healthy subjects and of diabetic patients considered as ulceration risky. Then, possible weight redistribution strategies are explored for each algorithm.