Sales Forecast In FMCG Industry Using Data Mining

Fast consumer goods are relatively cheap and fast selling products,which have the characteristics of short cycle and high frequency.Examples include non-durable goods,such as packaged food,drinks,cosmetics,non-prescription drugs and many other consumables.Different from durable consumables,consumers have low cost and multiple purchases,and the probability of irrational consumption is large.From a market perspective,it is characterized by high flux,low marginal profit,wide distribution network and high inventory turnover.According to these characteristics of fast consumer goods,using data mining and machine learning technology to carry out accurate sales prediction is of great significance for guiding all aspects of enterprise production,including purchasing production and inventory.This paper first proposes an analysis model for sales of commodity category.Aiming at commodity category level and commodity sales sequence formed in the past,the proposed analysis model aims at detecting the formed sequence,discerning the regularity of the sequence,analyzing the sequence of commodity aggregation to the category level,including potential explosive force model and seasonal quantitative model,respectively,the future trend and seasonal advance of commodity category.Conduct quantitative research.The explosive force model is used to study the potential explosive force and quantify the future explosive force of category commodities.Seasonal quantization model,based on the seasonal decomposition of sequence and signal-to-noise ratio,proposes seasonal quantitative algorithm to quantify the seasonal strength.For single commodity sales forecast,this paper uses different models for different types of commodities to classify and forecast.First,a sales forecasting method based on time series model is proposed.Fast moving consumer goods(FMCG)are a special kind of goods,which are very different from durable consumer goods.FMCG consumes a large amount,consumes quickly,consumes a short period of time,and many FMCG consumes seasonally.In view of these characteristics of fast moving consumer goods,this paper uses time series forecasting technology to analyze the seasonality of fast moving consumer goods,studies the sales forecasting model based on seasonal ARIMA time series,and proposes an automatic parameter adjustment method to achieve sales forecasting.Promotion is one of the important means to promote FMCG sales,FMCG is a high degree of homogeneity of the product,which determines the characteristics of FMCG Brand competition is the comprehensive competitions,channels,marketing,on the other hand,FMCG once has such as product selling price it is easy to obtain market opportunities.The fast consumption characteristics of FMCG itself decide that promotional activities play a decisive role in the sales of commodities.At present,all the fast selling companies are also doing all kinds of promotional activities.In this paper,a combination model of precision sales forecasting is proposed,which corresponds to the laws of the external variables and the historical sales of their own data.For historical sales data,a statistical analysis method based on time series is used to decompose the trend,seasonality and periodicity,and predict the future sales volume based on the improved time series model.At the same time,we use regression modeling to analyze the changing market information,business environment and external factors,and specifically analyze the complex interaction between marketing effect and external environment,and develop a combined model to predict future sales volume.Based on the above research results,this paper designs and implements a sales forecasting system for fast selling industry aiming at the demand of massive data.The system adopts B/S mode and multi-tier structure,and realizes two sales models and methods: no promotion goods and promotional products.This paper uses-the actual data of the enterprise to carry out experiments and tests.The results show that the accuracy of the two models achieved is 79% and 84% respectively.The forecast system runs smoothly and achieves the expected goal.