The goal of this research study is to develop and evaluate a modified genetic fuzzy system that will increase the algorithm’s efficiency and quality on small number of generation. The performance of the improved algorithm is carefully analyzed and assessed through multiple tests carried out on several datasets. In order to reduce the number of rules without decreasing classification accuracy and f-score, the modifications are examined. Two modifications with initialization and crossover operations and their combination with lexicase selection was checked on several datasets. Comparisons with alternative methods demonstrate that the proposed approach achieves comparable or slightly better accuracy in certain cases, indicating its potential as a viable alternative in addressing classification problems. The Mann-Whitney U test and careful statistical analysis support the importance of the changes.