Recent trends in fiber, yarnand fabric management in Textile-Apparel-Distribution network require a veryaccurate estimation of consumption pattern and sourcing management to minimizetheir costs and satisfy their customers. For such strategy, fashion designersrely more on fashion forecasting system to respond to the versatile textilemarket. However, the specific constraints of the textile consumption and sales(numerous and new items, short lifetime) complicate the forecasting procedureand distributors prefer to use intuitive estimation methods of the sales ratherthan the existing forecasting models. In this paper we firstly discuss the differentaspects used for improving the accuracy of fashion product development andforecasting process, use of computer technology in management of fiber, yarnand fabrics for estimation of their consumption. We also demonstrate a decisionaid system, based on neural networks, which automatically performs item salesforecasting of textile items with two models being demonstrated for silk fabricconsumption and cotton-textiles fabric consumption. Performances of our modelare evaluated using real data released by authorized agencies. Networkforecasted results show excellent correlation with real data emphasizing theits advantages over traditionally, statistical time series methods like movingaverage, auto-regression or combinations of them used for forecasting appareland fashion product sales.
Keywords:
Introduction
Introducing new products of Apparels made from different variations offibers, yarns and fabrics can provide a competitive advantage as well as along-term financial return on investment. New product development is thusvitally linked to a company's competitive strategy and consequently to the newproduct sales forecasting of apparels and textiles that provides quantitativeinformation on the
Traditionally, statisticaltime series methods like moving average, auto-regression or combinations ofthem are used for forecasting fashion trends and sales of fashion products.Since these models predict future sales only on the basis of previous sales,they fail in environments where sales are heavily influenced by exogenousvariables, such as economic conditions, climatic conditions, competitiveactivities, advertising and several factors. Although traditional statisticalmodels characterize the above mentioned factors, they are essentially linear innature and don't characterize the nonlinear nature of apparel sales and analysefibre, yarn, fabric consumption trends. Fuzzy logic, artificial neural networkscoupled with latest applications of computer technology for fashion design andmanagement and genetic algorithms(Gas) offer an alternative approach by takinginto account both endogenous and exogenous variables and allowing arbitrarynon-linear approximation functions derived(learned directly from the previousdata).
About the Authors
The authors are associatedwith Central Silk Technological Research Institute (CSTRI), Central Silk Board,Ministry of Textiles, Madivala, Bangalore, India and Department of Electronicsand Communication University Visveshvaraya College of Engineering (UVCE), Bangalore University K. R. Circle, Bangalore, India
Comments