关键词: 支持向量机回归, 混沌, 果蝇算法, 参数优化, 销售预测

Abstract: Aiming at the product sale time series of manufacturing enterprise with multi-dimension, small samples, nonlinearity, multi-peak, etc., a product sale forecasting method  based on the chaotic fruit fly support vector regression is presented. Chaos mapping theory is combined with fruit fly optimization algorithm, thereby increasing fruit fly population diversity and search ergodicity, and chaos perturbation is added in the optimization process, which avoids the search being trapped in local optimum and increases the ability to search for feasible solutions. This paper verifies optimal performance of Chaos Fruit Fly Optimization Algorithm（CFOA） by examples, then applies CFOA to optimize parameters of Support Vector Regression（SVR）, and builds sales forecasting model. The results of application show this model in sales prediction of auto parts is effective and feasible.

Key words: Support Vector Regression（SVR）, chaos, fruit fly optimization algorithm, parameter optimization, sales forecasting