This paper introduces the Probability-based Local Mean K-Nearest Neighbors (PLMKNN) algorithm, designed to surpass the limitations inherent in traditional K-Nearest Neighbors (KNN) and Local Mean K-Nearest Neighbors (LMKNN) classification approaches. Through adept feature weight computations, this methodology facilitates sophisticated optimization in feature selection and utilization. Post feature categorization, the LMKNN algorithm calculates class-specific probabilities, which is multiplied by pre-established weights, and yielded precise classification predictions. Empirical assessments utilizing the UCI dataset underscore the PLMKNN's substantial performance advantages over other KNN-based strategies, indicating a promising avenue for pioneering developments in the data classification sector.